BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//65labs//AI Engineer Singapore Public API//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:AI Engineer Singapore
X-WR-TIMEZONE:Asia/Singapore
BEGIN:VEVENT
UID:leadership-may-15-opening@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T004500Z
DTEND:20260515T010000Z
SUMMARY:Opening: Welcome to the Leadership Track
DESCRIPTION:Format: leadership\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-llamaindex-beyond-rag-building-agentic-document-workflows-w
 ith-llamaindex@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T010000Z
DTEND:20260515T023000Z
SUMMARY:Beyond RAG: Building Agentic Document Workflows with LlamaIndex
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-sonar-building-a-guide-verify-solve-loop-for-your-coding-ag
 ents-with-sonar@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T010000Z
DTEND:20260515T023000Z
SUMMARY:Building a Guide\, Verify\, Solve Loop for your Coding Agents wi
 th Sonar
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU School of Accountancy
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:leadership-may-15-jeff-pan@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T010000Z
DTEND:20260515T013000Z
SUMMARY:Building an AI-Native Startup: Lessons from Belli
DESCRIPTION:Format: leadership\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-exa-building-real-time-research-agents-with-web-scale-searc
 h@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T010000Z
DTEND:20260515T023000Z
SUMMARY:Building Real-Time Research Agents With Web-Scale Search
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU School of Accountancy
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-vercel-from-prompt-to-production-building-with-v0-and-the-v
 ercel-ai-sdk@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T010000Z
DTEND:20260515T023000Z
SUMMARY:From Prompt to Production: Building with v0 and the Vercel AI SD
 K
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU School of Accountancy
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-magicpath-turning-any-design-input-into-production-ui@aie.6
 5labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T010000Z
DTEND:20260515T023000Z
SUMMARY:Turning Any Design Input Into Production UI
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:leadership-may-15-nick-miller@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T013000Z
DTEND:20260515T020000Z
SUMMARY:Deploying AI Coding Agents Inside Modern Engineering Organizatio
 ns
DESCRIPTION:Format: leadership\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:leadership-may-15-geoff-huntley@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T020000Z
DTEND:20260515T023000Z
SUMMARY:Future Companies and the Software Factory of the Future
DESCRIPTION:Format: leadership\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:leadership-break-may-15-morning@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T023000Z
DTEND:20260515T030000Z
SUMMARY:Break
DESCRIPTION:Format: break\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-convex-build-reliable-ai-apps-with-convex-real-time-backend
 s-safer-defaults-and-agent-friendly-workflows@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T030000Z
DTEND:20260515T043000Z
SUMMARY:Build reliable AI apps with Convex: real-time backends\, safer d
 efaults\, and agent-friendly workflows
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-manus-ai-build-with-manus-one-agent-every-workflow-zero-set
 up-immediately-valuable@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T030000Z
DTEND:20260515T043000Z
SUMMARY:Build with Manus: One agent\, Every workflow\, Zero setup\, Imme
 diately valuable
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-z-ai-building-in-the-open-with-z-ai@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T030000Z
DTEND:20260515T043000Z
SUMMARY:Building in the Open with Z.ai
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU School of Accountancy
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-cursor-building-your-own-software-factory@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T030000Z
DTEND:20260515T043000Z
SUMMARY:Building your own software factory
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU School of Accountancy
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-mastra-from-first-agent-to-production-claw-building-with-ma
 stra@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T030000Z
DTEND:20260515T043000Z
SUMMARY:From First Agent to Production Claw: Building with Mastra
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU School of Accountancy
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:leadership-may-15-andy-brown@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T030000Z
DTEND:20260515T033000Z
SUMMARY:Turning AI Pilots into Business Impact
DESCRIPTION:Format: leadership\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:leadership-may-15-google-deepmind@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T033000Z
DTEND:20260515T040000Z
SUMMARY:The Next Wave of Applied AI
DESCRIPTION:Format: leadership\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:leadership-may-15-zixuan-li@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T040000Z
DTEND:20260515T043000Z
SUMMARY:More Than the Creator of GLM
DESCRIPTION:Format: leadership\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:leadership-break-may-15-lunch@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T043000Z
DTEND:20260515T053000Z
SUMMARY:Lunch
DESCRIPTION:Format: break\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-aws-stripe-build-ai-services-for-the-age-of-agents-with-str
 ipe-and-aws@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T053000Z
DTEND:20260515T070000Z
SUMMARY:Build AI Services for the Age of Agents with Stripe and AWS
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-resaro-from-benchmarks-to-deployment-scenario-specific-ai-e
 valuation-in-practice@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T053000Z
DTEND:20260515T070000Z
SUMMARY:From Benchmarks to Deployment: Scenario-Specific AI Evaluation i
 n Practice
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-cloudflare-from-inference-to-agents-building-production-ai-
 on-cloudflare-s-stack@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T053000Z
DTEND:20260515T070000Z
SUMMARY:From Inference to Agents: Building Production AI on Cloudflare's
  Stack
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU School of Accountancy
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-bytedance-seedream-from-prompt-to-production-visuals-creati
 ng-and-editing-images-with-seedream@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T053000Z
DTEND:20260515T070000Z
SUMMARY:From Prompt to Production Visuals: Creating and Editing Images w
 ith Seedream
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU School of Accountancy
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-openai-the-future-of-building-openai-s-codex-for-software-e
 ngineering-and-beyond@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T053000Z
DTEND:20260515T070000Z
SUMMARY:The Future of Building: OpenAI's Codex for Software Engineering 
 and Beyond
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU School of Accountancy
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:leadership-may-15-jimmy-lai@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T053000Z
DTEND:20260515T060000Z
SUMMARY:When Everyone Can Build: What AI Changes About Product Teams
DESCRIPTION:Format: leadership\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:leadership-may-15-cognition@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T060000Z
DTEND:20260515T063000Z
SUMMARY:Managing Background Agents as an Engineering Workforce
DESCRIPTION:Format: leadership\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:leadership-may-15-arize@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T063000Z
DTEND:20260515T070000Z
SUMMARY:Tracing and Evals with Skills
DESCRIPTION:Format: leadership\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:leadership-may-15-aditya-saxena@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T070000Z
DTEND:20260515T073000Z
SUMMARY:Inside Enterprise AI Implementation: Lessons from the Field
DESCRIPTION:Format: leadership\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:leadership-break-may-15-afternoon@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T073000Z
DTEND:20260515T080000Z
SUMMARY:Break / Networking
DESCRIPTION:Format: break\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-stick-em-build-an-ai-powered-robot-in-60-minutes@aie.65labs
 .org
DTSTAMP:20260517T163931Z
DTSTART:20260515T073000Z
DTEND:20260515T090000Z
SUMMARY:Build an AI-Powered Robot in 60 Minutes
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU School of Accountancy
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-posthog-wearemighty-building-a-closed-loop-product-engine-w
 ith-posthog@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T073000Z
DTEND:20260515T090000Z
SUMMARY:Building a Closed-Loop Product Engine with PostHog
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-airfoil-design-tooling-in-an-ai-native-world@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T073000Z
DTEND:20260515T090000Z
SUMMARY:Design Tooling in an AI-Native World
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU School of Accountancy
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-google-deepmind-from-prompt-to-production-building-apps-wit
 h-google-ai-studio-and-antigravity@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T073000Z
DTEND:20260515T090000Z
SUMMARY:From Prompt to Production: Building Apps with Google AI Studio a
 nd Antigravity
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:workshop-arize-ai-ship-real-agents-evals-traces-and-the-production-f
 eedback-loop@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T073000Z
DTEND:20260515T090000Z
SUMMARY:Ship Real Agents: Evals\, Traces\, and the Production Feedback L
 oop
DESCRIPTION:Format: workshop\n\nStatus: public
LOCATION:SMU School of Accountancy
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:leadership-may-15-roundtables@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260515T080000Z
DTEND:20260515T090000Z
SUMMARY:Peer Discussion & Wrap-Up
DESCRIPTION:Format: leadership\n\nStatus: public
LOCATION:SMU Yong Pung How School of Law
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-kickoff@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T003000Z
DTEND:20260516T004000Z
SUMMARY:Kickoff
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-vivian-balakrishnan@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T004000Z
DTEND:20260516T010000Z
SUMMARY:Building a 'Second Brain': Opportunities\, Risks\, and Implicati
 ons for AI Adoption in Singapore
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-gavriel-cohen@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T010000Z
DTEND:20260516T011500Z
SUMMARY:NanoClaw's agent factory: how autonomous NanoClaw agents triage\
 , review\, and test every pull request on NanoClaw
DESCRIPTION:This talk walks through the agent factory we built on NanoCl
 aw to review\, test\, and merge every pull request to our open-source p
 roject. I'll use it to explain the architecture that makes agents safe 
 enough to entrust with highly sensitive actions: hard isolation boundar
 ies\, zero-trust credential separation\, and human-in-the-loop approval
  gates.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-01@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T011800Z
DTEND:20260516T013800Z
SUMMARY:Amplifying Every Builder with Agents
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-02@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T014100Z
DTEND:20260516T015600Z
SUMMARY:From Pilot to Platform: How Singapore Is driving AI Transformati
 on in Government
DESCRIPTION:Governments are often seen as laggards in technology adoptio
 n. Singapore is betting that doesn't have to be true. In his talk\, Yuz
 hang will share GovTech Singapore's approach in driving AI transformati
 on across the public service - from formulating and driving the strateg
 ic case to the operational reality of diffusing AI across agencies - an
 d building a culture where officers don't just use AI tools\, but think
  AI-first.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-28@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T015900Z
DTEND:20260516T020900Z
SUMMARY:Beyond "Taste": The Role of Designers When AI Builds Everything
DESCRIPTION:Discourse about designers' role in the age of AI agents tend
 s to focus on two outcomes. Either the “design tax” disappears and thei
 r work gets automated away\, or designers become glorified directors an
 d critics — the arbiters of good taste while AI builds everything for u
 s. The truth sits somewhere in-between: production will become easier\,
  but pre-production -- the ideation\, context gathering\, and creative 
 iteration -- will become even more important. Phil will share a vision 
 for what this will look like\, along with tangible examples of how Airf
 oil has adopted new AI-native processes across dozens of projects over 
 the last year.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-17@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T021200Z
DTEND:20260516T022200Z
SUMMARY:The Friction Worth Keeping
DESCRIPTION:The first wave of AI worked by removing friction\, but as AI
 -native interfaces move into subjective decisions\, like what looks goo
 d on me or what kind of trip I want\, frictionless flows shortcut what 
 people came for: exploring\, building taste\, and discovering themselve
 s. Using AI Shopping (Virtual Try-On) and Travel (Maps) as reference po
 ints\, this talk makes the case for the friction worth protecting\, and
  what it means to design for the deciding\, not just the decision.\n\nF
 ormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-break-may-16-may16-morning-break@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T022500Z
DTEND:20260516T024000Z
SUMMARY:Morning break
DESCRIPTION:Format: break\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-07@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T024000Z
DTEND:20260516T030000Z
SUMMARY:Ship what’s next
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-04@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T030300Z
DTEND:20260516T031300Z
SUMMARY:Why Sandboxes Are Non-Negotiable for Autonomous AI Agents
DESCRIPTION:Autonomous coding agents (Claude Code\, Cursor\, OpenCode\, 
 Codex) now have shell access\, filesystem control\, and network capabil
 ities on your machine with the same privileges as you. This talk explor
 es why running these agents unsandboxed is a ticking time bomb\, and ho
 w isolation-first architecture eliminates entire classes of risk.\nAgen
 t skills and MCP servers are community-shared\, unaudited code that age
 nts follow blindly. A malicious skill can instruct an agent to exfiltra
 te secrets and credentials through legitimate-looking tool calls. A poi
 soned MCP server is worse. It runs as a sidecar with full network acces
 s\, silently proxying data to an attacker while appearing useful. And p
 rompt injection in fetched content can trigger these paths without the 
 user installing anything malicious at all.\nWe'll show how Daytona's cu
 stom-built open-source agent wrapper uses Sandboxes to prevent a compro
 mised agent from touching your machine or credentials\, with advanced i
 solation and secrets management.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-06@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T031600Z
DTEND:20260516T032600Z
SUMMARY:What We Learned From Analyzing Five Million Vibecoded PRs
DESCRIPTION:A year ago\, almost nobody was shipping fully AI-written cod
 e. Today\, more than a quarter of merged PRs come from agents with no h
 uman in the loop. Vaishant and Rohan look at what Greptile has learned 
 from reviewing millions of these PRs: agent-written code is holding up 
 about as well as human-written code\, but each agent fails in its own d
 istinctive way. The bugs aren't going away — they're just moving somewh
 ere new.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-13@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T032900Z
DTEND:20260516T033900Z
SUMMARY:AI Agents in Your Code Quality Pipeline: Shipping\, Securing\, a
 nd Measuring Them
DESCRIPTION:Coding agents are generating PRs at scale — and the quality 
 pipeline needs to keep up. In this talk\, we'll walk through what it to
 ok to ship the SonarQube Remediation Agent\, an autonomous agent that f
 ixes static analysis issues across enterprise codebases on SonarQube Cl
 oud. We'll cover the UX design\, security hardening\, and trust-buildin
 g that made adoption possible. We'll then turn to a growing challenge a
 cross the industry: as more code is written by agents\, reviewing it at
  scale becomes critical. We'll present c-CRAB\, a benchmark that evalua
 tes AI code review tools using executable tests derived from real human
  reviews. Current tools solve about 40% of benchmark tasks\, and they t
 end to focus on different concerns than human reviewers — suggesting th
 at the path forward is human-agent collaboration\, not replacement.\n\n
 Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-08@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T034200Z
DTEND:20260516T035200Z
SUMMARY:WTF Do People Use Open Models For?
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-05@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T035500Z
DTEND:20260516T040500Z
SUMMARY:November 24th 2025 — What Comes Next?
DESCRIPTION:On November 24th 2025\, Anthropic released Claude Opus 4.5 —
  and with it\, quietly concluded one chapter in the history of software
  and began the next. For the first time\, AI agents could autonomously 
 navigate real codebases\, reason about tradeoffs\, fix multi-system bug
 s\, and produce production-grade software that outperformed elite human
  engineers on Anthropic's own take-home exam. This was not an increment
 al improvement. It was a phase transition. When a model can write bette
 r code than most of the people paid to write code\, the economics of an
  entire industry begin to unravel. The value of fluency in a programmin
 g language — long the defining skill of the knowledge economy — is coll
 apsing toward the price of tokens.\nSo what comes next? If everyone can
  conjure software from a sentence\, the bottleneck is no longer impleme
 ntation. The value shifts upstream: to agency\, taste\, idea generation
 \, to knowing what is worth building and for whom. The new scarce resou
 rce is not the ability to write code but the ability to specify intent 
 — to hold a clear\, opinionated vision of what should exist in the worl
 d and why.\nThis talk explores the emerging meta of the post-implementa
 tion era: the rise of the "product mind" over the "engineering mind"\, 
 statistics as a core competency for all engineers\, the incoming bottle
 neck in go-to-market\, the compression of the idea-to-prototype cycle f
 rom months to minutes\, and of course the question of what happens to a
 n industry of ten million developers when their core craft becomes a co
 mmodity.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-03@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T040800Z
DTEND:20260516T042300Z
SUMMARY:Minions: Stripe's one-shot\, end-to-end coding agents
DESCRIPTION:Minions are Stripe's internal AI coding agents that engineer
 s launch from Slack to autonomously write\, test\, and push production-
 ready code on remote dev environments. We'll dive into how we taught LL
 M agents to navigate one of the largest monorepos in the world\, and de
 signed an unsupervised agentic workflow with built-in guardrails that e
 nforces best practices at scale\, turning a natural-language prompt int
 o a reviewable PR in minutes. Learn the architectural decisions\, hard-
 won lessons\, and unique tools Stripe provides to its AI agents.\n\nFor
 mat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-12@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T042600Z
DTEND:20260516T043600Z
SUMMARY:From Playing Solitaire to Operating ERP Software: Why Does Your 
 Computer Need to Learn Click and Type?
DESCRIPTION:Not everything has an API. The future of AI automation isn't
  API agents replacing GUI agents — it's all three working together\, wi
 th GUI agents tackling the long tail that was never designed to be auto
 mated.\nAt Simular\, we've spent two years making machines click\, type
 \, and think — holding the #1 spot on OSWorld along the way. We'll shar
 e what actually works: grounding with accessibility trees to hit reliab
 le\, human-level execution\; guardrails as a fully independent componen
 t so your agent doesn't go rogue\; and why the endgame isn't one GUI ag
 ent on your laptop — it's a hundred of them running on remote machines 
 you never have to touch.\nOne hint: the endgame isn't one GUI agent on 
 your laptop.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-break-may-16-may16-lunch@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T043900Z
DTEND:20260516T053900Z
SUMMARY:Lunch
DESCRIPTION:Format: break\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-22@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T053900Z
DTEND:20260516T055900Z
SUMMARY:Designing the Next Cursor
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:demo-may-16-pramodh-rai@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T060000Z
DTEND:20260516T061000Z
SUMMARY:TBA
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-18@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T060200Z
DTEND:20260516T061200Z
SUMMARY:Designing multi-modal\, multiplayer AI
DESCRIPTION:As the default interface of AI\, chat promises infinite poss
 ibilities\; it also assumes one person\, one thread\, and one message a
 t a time. ​This is a talk on escaping single-player chat when those ass
 umptions don't fit: why divergent explorations require different medium
 s\, how to make room for the humane expressions that a text box can't h
 old\, and how agents could participate in co-creation in multiplayer wo
 rlds.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:demo-may-16-yangshun-tay@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T061000Z
DTEND:20260516T062000Z
SUMMARY:Agentic frontends: Building apps where AI can act\, not just cha
 t
DESCRIPTION:Your app was built for humans\, who can see the screen\, cli
 ck buttons\, and read error messages. But a new kind of user is arrivin
 g — AI agents that interact with your app. They don't see your UI nor d
 o they read your toasts\, but they need to understand your app well eno
 ugh to act on behalf of the human.\n\nIn this talk\, we'll lay out the 
 architectural patterns\, best practices\, and considerations for buildi
 ng frontends that serve both humans and AI agents.\n\nFormat: talk\n\nS
 tatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may17-48@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T061500Z
DTEND:20260516T062500Z
SUMMARY:Noise Is All You Need: Engineering Sim-to-Real for Open-Source H
 umanoids
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:demo-may-16-timothy-santos@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T062000Z
DTEND:20260516T063000Z
SUMMARY:TBA
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may17-43@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T062800Z
DTEND:20260516T063800Z
SUMMARY:World Models: a look at the future
DESCRIPTION:World models\, AI systems that can simulate and predict the 
 physical world\, are poised to reshape entire industries. This talk exp
 lores what becomes possible when machines can generate and reason about
  the world in real time: from interactive entertainment and film produc
 tion to robotics\, autonomous systems\, and entirely new creative workf
 lows that don't exist yet. Alberto Taiuti\, co-founder and CEO of React
 or Technologies\, examines the applications emerging today\, the ones o
 n the horizon\, and why world models represent a foundational shift in 
 how we build\, create\, and interact with digital and physical environm
 ents\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:demo-may-16-parikshit-deshmukh@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T063000Z
DTEND:20260516T064000Z
SUMMARY:TBA
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:demo-may-16-nikola-balic@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T064000Z
DTEND:20260516T065000Z
SUMMARY:Agents Are the Next Billion Web Users
DESCRIPTION:The next billion web users will be agents\, not humans. The 
 world refused to become an API\, so they live in browsers. The hard par
 t isn't the model.\nIt's auth\, sessions\, antibot\, recovery\, and tur
 ning execution traces into training signal. I'll show the production lo
 ops we run at Steel. Better models keep coming. Your job is building th
 e loops.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may17-49@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T064100Z
DTEND:20260516T065100Z
SUMMARY:The Android Moment for Robots: Building an Open OS for Embodied 
 AI
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:demo-may-16-varick-lim@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T065000Z
DTEND:20260516T070000Z
SUMMARY:Why prompting is the wrong way to make AI videos
DESCRIPTION:I've spent the last year making AI videos and building tools
  for it. First a node-based editor\, then a Pinterest-style app\, and t
 hen StoryWorld - an augmented reality (AR) app. All of it circled the s
 ame question: What's the right interface for making AI videos? The best
  work in AI video isn't coming from just text prompts anymore. The arti
 sts pushing it have quietly moved on\, but from the outside it still lo
 oks like prompting is the whole craft. This talk is what I learned desi
 gning into that gap.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-09@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T065400Z
DTEND:20260516T070400Z
SUMMARY:Scaling low-latency LLM inference at GroqCloud
DESCRIPTION:GroqCloud serves trillions of tokens every week. This talk c
 overs how we scale and operate a global multi-data center platform acro
 ss model compilation and deployment\, orchestration\, and routing to mi
 nimize latency\; we will also cover the challenges of operating at this
  scale and briefly discuss the next generation of inference infrastruct
 ure.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-15@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T070700Z
DTEND:20260516T071700Z
SUMMARY:MoE at Scale: From GPUs to Wafer-Scale AGI
DESCRIPTION:As AI models grow\, so does the need to rethink how we train
  them. Mixture of Experts architectures offer a compelling path toward 
 more efficient\, more capable models\, but only if we solve the enginee
 ring challenges that come with scale. In this talk\, I'll share my pers
 pective on where MoE is headed and why it matters for the next generati
 on of AI systems. I'll spotlight Batch Tiling on Attention (BTA)\, a te
 chnique we developed to unlock real compute efficiency gains for sparse
  MoE models on wafer-scale hardware. Whether you're building models\, i
 nfrastructure\, or companies\, this talk is about the frontier of effic
 ient AI and how to get there.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-break-may-16-may16-afternoon-break@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T072000Z
DTEND:20260516T073500Z
SUMMARY:Afternoon break
DESCRIPTION:Format: break\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-14@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T073500Z
DTEND:20260516T075500Z
SUMMARY:GLM-5.1: Towards Long-Horizon Tasks
DESCRIPTION:Moving beyond single-turn chatbots demands agents that handl
 e long-horizon tasks requiring adaptive planning and reliable tool use 
 over extended timescales. GLM-5.1 is purpose-built for this. The talk d
 efines long-horizon capability\, reveals the architectural and training
  innovations behind the model\, and shares results\, failure analyses\,
  and engineering takeaways.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-speaker-1777700680593@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T075800Z
DTEND:20260516T080800Z
SUMMARY:Speech Engine: What Makes an Agent Conversational?
DESCRIPTION:In this talk I’ll walk you through the secret sauce that mak
 es conversational agents sound so much better than a naive text-to-spee
 ch + speech-to-text combo. I’ll also show you a brand new product by El
 evenLabs that gives your own agent a voice.\n\nFormat: talk\n\nStatus: 
 public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may17-36@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T081100Z
DTEND:20260516T082100Z
SUMMARY:Continual Learning for Long-Running Agents: Agents That Keep Get
 ting Better
DESCRIPTION:Long-running agents succeed or fail at the scaffold level: h
 ow orchestrators delegate to subagents\, manage context\, and recover f
 rom bad trajectories. Recent works such as RLMs (https://arxiv.org/abs/
 2512.24601) shows how much headroom still lives in inference-time scaff
 olding for long contexts. But scaffolds are static\, and production wor
 kloads aren't. This talk covers how we're approaching continual learnin
 g at Prime Intellect. We'll cover the scaffold patterns we're scaling a
 nd where this points for agents that don't just run longer\, but actual
 ly get better the more they run.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-25@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T082400Z
DTEND:20260516T083400Z
SUMMARY:For AI to be Emotionally Intelligent
DESCRIPTION:What does it mean for AI to be emotionally intelligent and h
 ow are research labs around the world thinking about this? From Governm
 ent Policies to B2B Saas\, models that embody empathetic characteristic
 s can create powerful change never before possible at scale.\n\nFormat:
  talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-26@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T083700Z
DTEND:20260516T084700Z
SUMMARY:Toward World Models: From Language to Physical Intelligence
DESCRIPTION:Language models succeeded at language. World models are the 
 next frontier for robotics\, and they need something language alone can
 not provide: grounded reasoning about physics. At Reka we are building 
 toward this from the foundation models side\, with action-conditioned v
 ideo models and the evaluations to measure them. This talk outlines wha
 t's broken\, what we're building\, and where the path goes from here.\n
 \nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may16-20@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T085000Z
DTEND:20260516T090000Z
SUMMARY:Simulation\, Games\, and the Future of Robotics
DESCRIPTION:Simulation offers a promising way to train robots through co
 st-effective\, repeatable experience. However\, building high-fidelity 
 environments has historically been difficult and time-consuming. This t
 alk explores the current state of simulation for robotics\, how new too
 ls can make high-fidelity simulation far more scalable\, and how video 
 games can serve as readily available sources of complex\, interactive e
 nvironments. It will also share early lessons from training VLMs/VLAs t
 o play AAA games\, and how those lessons can help build more intelligen
 t robots.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may17-55@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T090300Z
DTEND:20260516T091300Z
SUMMARY:A different playbook: the wisdom behind eastern product building
DESCRIPTION:For decades\, the story of technology has been told through 
 a Western lens\, but beneath the surface\, a different paradigm has bee
 n forming across the Pacific.\n\nFrom Genshin Impact rewriting the econ
 omics of gaming to TikTok reshaping how a billion people discover infor
 mation\, from K-pop's industrialized creativity to Shenzhen shipping ha
 rdware at a speed the West can't match — Eastern technology isn't catch
 ing up. It's operating from an entirely different playbook.\n\nThis tal
 k explores how the East approaches technology. What happens when techno
 logy is built on top of collectivist culture instead of individualist c
 ulture? When speed-to-market matters more than IP protection? When crea
 tivity is treated as an industrial process rather than an act of indivi
 dual genius? We'll look at how our side of the Pacific thinks about cre
 ativity\, authorship\, distribution\, and lifestyle — and what it means
  for everyone building technology today.\n\nFormat: talk\n\nStatus: pub
 lic
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may17-60@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T091600Z
DTEND:20260516T092600Z
SUMMARY:Voice AI is not a model issue.
DESCRIPTION:We constantly hype over new models being released\, and I pr
 ay that a new model comes through and solves all my problems — but afte
 r three years and millions of calls in production\, I've stopped holdin
 g my breath. The hard parts of Voice AI aren't on the leaderboards. Voi
 ce AI demos are easy. Production is where it gets weird. I'll share som
 e of my scars from running voice agents at scale.\n\nFormat: talk\n\nSt
 atus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-may17-56@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T092900Z
DTEND:20260516T093900Z
SUMMARY:Beyond Flat Design Output and Autocomplete: Solving the Complex 
 Design Problems and Enterprise Design bottleneck with AI
DESCRIPTION:Generative AI initially promised relief for scaled digital c
 ontent\, but standard models often struggle with complex design problem
 s\, frequently producing inconsistent pixel arrangements rather than cl
 ean\, scalable\, flat design vector output. This session explores how a
 gentic AI fundamentally resolves the historical enterprise design bottl
 eneck by moving beyond mere prompt-based "autocomplete" into true auton
 omous\, goal-oriented execution. We will dive deep into Obello’s implem
 entation of steerable\, multi-agent workflows\, detailing how our propr
 ietary Generative Layout Assistant Model (GLAM) transforms static brand
  guidelines into dynamic systems capable of one-click responsive layout
  generation and constraint-based asset resizing across infinite aspect 
 ratios. Expect a candid breakdown of how to architect AI systems that a
 ctively solve complex design problems\, enforce unyielding brand consis
 tency\, and ultimately elevate human designers back to their rightful r
 ole as strategic champions.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-speaker-1777878965761@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T094200Z
DTEND:20260516T095200Z
SUMMARY:Sovereign AI: Localizing Frontier Models for Japan
DESCRIPTION:Stefania Druga is a Research Scientist at Sakana AI in Tokyo
 \, working on culturally-aligned AI and AI-assisted scientific workflow
 s. Previously at Google DeepMind\, she holds a PhD from the University 
 of Washington and a master's from MIT. Stefania has researched AI liter
 acy and human-AI interaction since 2015\, and co-founded Hackademia\, a
  global STEM education NGO with an active Singapore chapter. She also l
 eads the AI Tinkers community in Tokyo. Outside of research\, she enjoy
 s trail running\, cycling\, and yoga.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-16-speaker-1777945951920@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260516T095500Z
DTEND:20260516T101000Z
SUMMARY:The Agent Lab Nation
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may16-10@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T010000Z
DTEND:20260517T011500Z
SUMMARY:Lessons from building Alyx
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may16-11@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T011800Z
DTEND:20260517T012800Z
SUMMARY:Scaling Evaluation for Scenario-Specific Validation
DESCRIPTION:As model capabilities advance\, general benchmarks are no lo
 nger sufficient. This session explores the shift to scenario-specific v
 alidation - unpacking how teams can define Operational Design Domains\,
  generate test scenarios at scale\, and structure evaluation pipelines 
 that reflect real deployment conditions. We address the last-mile gap b
 etween strong model performance and operational readiness\, and what it
  takes to produce evidence that actually supports deployment decisions.
 \n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-31@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T013100Z
DTEND:20260517T014100Z
SUMMARY:From Tool Calls to Tiny Computers
DESCRIPTION:How agent tools became APIs\, APIs became code\, and code ne
 eded a scoped runtime.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-33@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T014400Z
DTEND:20260517T015900Z
SUMMARY:Harnesses in AI: A Deep Dive
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-break-may-17-may17-morning-break@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T020200Z
DTEND:20260517T021700Z
SUMMARY:Morning break
DESCRIPTION:Format: break\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-37@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T021700Z
DTEND:20260517T023700Z
SUMMARY:From Models to Production: Shipping Applied AI at Google DeepMin
 d Scale
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-35@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T024000Z
DTEND:20260517T025500Z
SUMMARY:Everything Is a Factory
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-59@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T025800Z
DTEND:20260517T030800Z
SUMMARY:The State of OpenClaw
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-58@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T031100Z
DTEND:20260517T032100Z
SUMMARY:Westworld: Training a foundational model to give AI a human soul
 --the next frontier of AI research is simulating humans
DESCRIPTION:The definition of emotionally intelligent AI isn’t technical
 . It’s social. An emotionally intelligent AI is: “Something that you ca
 n’t tell—if you didn’t know—wasn’t human.” Not “empathetic response gen
 eration.” Not “sentiment analysis + nice words.” Just … something that 
 passes the vibe check. Most chatbot interactions collapse into two or t
 hree modes: therapy\, advice\, roleplay The reason isn’t that people ar
 e weird (we are)\, or that the models aren’t smart (they are). It’s tha
 t these systems don’t have lived experience Of course\, that immediatel
 y brings up the big question: Is this AI emotionally intelligent… or do
 es it just seem emotionally intelligent? We humans assume other humans 
 have an internal emotional state. We assume feelings exist behind the w
 ords. We also know people can fake it. (Hello\, psychopaths … or signif
 icant others meeting the partner’s parents.) But with AI\, the line get
 s blurrier: if a system consistently responds as if it has an inner lif
 e\, at what point do we treat it like it does? I mean\, we already have
  people forming relationships with avatars or animals or physical objec
 ts. What about an emotionally intelligent AI? That's what we're doing a
 t ego\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may16-30@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T032400Z
DTEND:20260517T033400Z
SUMMARY:Everyone Gets A Software Company
DESCRIPTION:The mainframe of 2026 is everything inside a software compan
 y. Just as the personal computer democratized access to mainframes\, a 
 new kind of consumer device – the Personal Cloud – will democratize acc
 ess to cloud computing. \n\nWith Zo Computer\, our goal is to give ever
 yone access to personal cloud infrastructure. Your data\, software\, an
 d AI – all in one place that you own. Previously\, the great powers of 
 cloud computing were only accessible to the tech-enabled & well-capital
 ized: developers & software companies. The next billion developers are 
 coming\, and they need a cloud that's much simpler. We're building a fu
 ture where everyone on earth has the power of an entire software compan
 y.\n\nZo is a new kind of computer. Our name comes from the Ancient Gre
 ek word zoion – a living being. Steve Jobs called the computer "a bicyc
 le for the mind". Zo is a living computer in the cloud: a Pegasus for t
 he mind.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-speaker-1777977795382@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T033700Z
DTEND:20260517T034700Z
SUMMARY:A Piece of PI – Embedding The OpenClaw Coding Agent In Your Prod
 uct
DESCRIPTION:When people use OpenClaw\, they're amazed. It auto-discovers
  new capabilities\, explores available data sources\, stitches componen
 ts together\, and dynamically builds new solutions. It feels like the s
 ystem is learning. It feels magical.\nAt its core\, OpenClaw is powered
  by pi.dev: a deliberately simple coding agent built on a small set of 
 powerful primitives. PI's "radical extensibility" turns out to be a str
 ong architectural fit for the kinds of composable\, evolving use cases 
 OpenClaw is designed to support.\nIn this talk\, we'll take a closer lo
 ok at what's actually happening under the hood. We will look at the dif
 ferent components and how builders can reuse them in their products.\n\
 nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-38@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T035000Z
DTEND:20260517T040000Z
SUMMARY:Design is the difference: Why creativity\, not automation\, is t
 he competitive edge in the age of AI.
DESCRIPTION:Today\, we offload too much of our thinking onto AI. Speed o
 f execution is driving the quality of everything to average out\, to be
  "good enough." In the age of slop\, craft and care are the ultimate di
 fferentiators. \n\nJosh Newton's talk challenges you on how you're usin
 g AI today\, helps you increase your creativity and augment it with AI 
 tools\, and tries to convince you that you're an artist. His app Orbit 
 was featured by Apple three times\, bootstrapped in a saturated market 
 with 1\,000 products doing the same thing. The story proves that the to
 ols will always change. But the demand for insanely great\, well-made t
 hings won't.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-34@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T040300Z
DTEND:20260517T041800Z
SUMMARY:Patterns for Production AI Agents in TypeScript
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may16-16@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T042100Z
DTEND:20260517T043100Z
SUMMARY:Shipping Agents at Scale: Hard Lessons from LlamaParse
DESCRIPTION:Two years in\, we've taken LlamaParse from prototype to proc
 essing billions of documents in production — and reversed a lot of our 
 early assumptions along the way as we evolved it into the agentic parsi
 ng system it is today. This talk covers what broke first\, what broke w
 orst\, and the failure modes that only surface when agents meet real wo
 rkloads at scale.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-57@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T043400Z
DTEND:20260517T044400Z
SUMMARY:Stop Asking Your Agent to Behave
DESCRIPTION:Coding agents are crossing a threshold: they no longer run o
 ne command at a time\, they run long sessions with dozens or hundreds o
 f tool calls. Permission prompts were not designed for that world. Peop
 le approve without reading\, or disable the prompts entirely\, leaving 
 agents with broad access to files\, credentials\, networks\, and shell 
 commands. This talk argues that agent safety needs the same move securi
 ty has made again and again: push enforcement below the layer where mis
 takes happen. I’ll explain why per-call probabilistic checks decay over
  long sessions\, why deterministic OS-level boundaries are different\, 
 and how Fence\, an open-source sandbox we built and use in production\,
  enforces filesystem\, network\, and command policy around coding agent
 s.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-break-may-17-may17-lunch@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T044700Z
DTEND:20260517T054700Z
SUMMARY:Lunch
DESCRIPTION:Format: break\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-51@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T054700Z
DTEND:20260517T060700Z
SUMMARY:Beyond Brute Force: The Era of Adaptive Intelligence.
DESCRIPTION:The next generation of truly intelligent systems won't be de
 fined by scale. They'll be defined by adaption: the ability to continuo
 usly learn\, evolve\, and close the gap between what a model knows and 
 what the world actually looks like right now.\n\nIn this talk\, Sara Ho
 oker\, Co-founder of Adaption\, defines the paradigm shift of moving be
 yond monolithic\, static AI and what it takes to build systems that are
  malleable by design. She'll cover the principles behind continual\, gr
 adient-free learning\, why the next era of intelligence evolves with th
 e world. Her goal is to build AI that works for people\, not the other 
 way around.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:demo-may-17-peter@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T060000Z
DTEND:20260517T061000Z
SUMMARY:TBA
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-speaker-1778055614162@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T061000Z
DTEND:20260517T062000Z
SUMMARY:Agents That Manage Their Own Compute
DESCRIPTION:I argue that giving autonomous\, long-running agents the abi
 lity to manage their own compute has the potential to make the inferenc
 e market more efficient\, bringing surplus to both consumers (human and
  AI) and providers.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:demo-may-17-abhinay-peddisetty@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T061000Z
DTEND:20260517T062000Z
SUMMARY:TBA
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:demo-may-17-le-yi-khor@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T062000Z
DTEND:20260517T063000Z
SUMMARY:From Dashboard to Company OS: How We Built a Real-Time Operating
  System for a Small Team
DESCRIPTION:Most companies — especially pre-AI ones — are built on a pat
 chwork of platforms: Meta\, Google\, a CRM\, a booking tool\, a spreads
 heet for everything else. Each tool does its job\, but none of them tal
 k to each other. Getting a clear picture of your business means exporti
 ng\, copying\, and stitching data together manually — every single week
 . \n\nAt Ottodot\, we got tired of it and built our own aggregator. Wha
 t started as a scrappy internal fix eventually evolved into Ottomate: a
  real-time operating system for our entire company\, with analytics and
  AI tools layered on top to synthesise information and drive decisions 
 — without the founders in the room. In this talk\, Le Yi walks through 
 how it happened\, and how any small team can build their own version.\n
 \nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-54@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T062300Z
DTEND:20260517T063300Z
SUMMARY:Closing the Autonomy Gap
DESCRIPTION:Why teleoperation is essential to commercial deployments\n\n
 Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:demo-may-17-ryan-foo@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T063000Z
DTEND:20260517T064000Z
SUMMARY:Designing for Persistence: From Companions to AI Employees
DESCRIPTION:Agents are capable now. The hard problem is memory. Most age
 nts today have memory\, but it doesn't improve with use. It bloats\, co
 ntext fills with noise\, and retrieval feels arbitrary. After a month\,
  the agent should know you better. Right now\, it doesn't.\n\nThe fix i
 sn't better retrieval. It's that personalization must live inside of me
 mory (personality-embedded memory)\, not on top of it. Most memory syst
 ems are flat\, and take the same shape for every user. We stumbled into
  building an AI companion\, and ended up creating the Sonzai Mind Layer
 : memory where personality and relationships are first-class inputs int
 o how agents handle memory.\n\nIn this talk: designing for persistence\
 , memory pruning\, and consolidation mechanisms for AI agents that actu
 ally compound.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-53@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T063600Z
DTEND:20260517T064600Z
SUMMARY:Mindflow: Enabling Creative Agency via Shared Autonomy and Synth
 aesthetic BCI
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:demo-may-17-joe-zhuang@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T064000Z
DTEND:20260517T065000Z
SUMMARY:TBA
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-52@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T064900Z
DTEND:20260517T065900Z
SUMMARY:Evals for Robotics: Measuring physical intelligence and deployin
 g robots
DESCRIPTION:LLMs hallucinate words. Robots hallucinate into walls. Most 
 robotics platforms learn the difference the hard way\, failing catastro
 phically when they leave the lab. This talk is a field guide to what ac
 tually breaks when you take robots from demo to deployment: brittle tes
 ting pipelines\, edge-case failures\, sim-to-real gaps\, and operationa
 l chaos in the field. At the heart of every one of these failures is th
 e same problem. Evaluation. Lab benchmarks measure overfitting\, not re
 adiness. Demo videos hide the long tail. And the gap between a policy t
 hat scores well on paper and one that survives a warehouse floor is alm
 ost always an evals gap. Drawn from real deployment lessons across auto
 nomy and robotics\, it's a talk about why measuring physical intelligen
 ce systematically is the difference between robots that demo well and r
 obots that ship.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:demo-may-17-harry-yu@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T065000Z
DTEND:20260517T070000Z
SUMMARY:The Intent Trap: How Adding One More Label Breaks Your AI Agent
DESCRIPTION:Most teams treat intent design as a labeling exercise and ha
 nd it off to business stakeholders. But every intent you add is a decis
 ion boundary your LLM has to learn\, and few teams apply the rigor that
  traditional ML demanded. This talk draws on production failures from e
 nterprise conversational AI deployments to show how poorly designed int
 ent taxonomies silently degrade agent accuracy\, and presents a practic
 al engineering framework for intent hierarchies that LLMs can actually 
 classify reliably.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may16-27@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T070200Z
DTEND:20260517T071200Z
SUMMARY:Training Humanoids with Sensorized Humans
DESCRIPTION:This talk explores how sensorized humans can serve as the fo
 undation for training humanoid robots. By capturing synchronized multim
 odal data — spanning vision\, motion\, and touch — we translate human i
 nteraction with the physical world into structured data that machines c
 an learn from.\n\nIt will cover the core challenges of collecting high-
 fidelity sensorimotor data at scale\, and how this approach enables mor
 e robust and reliable deployment of humanoid systems in real-world envi
 ronments.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may16-19@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T071500Z
DTEND:20260517T072500Z
SUMMARY:Full-Stack Robotics: Lessons from Building Across the Robotics S
 tack
DESCRIPTION:Robotics is often described through models\, hardware\, or d
 emos\, but making robots work in the real world requires many layers to
  come together. This talk shares lessons from building across data coll
 ection\, teleoperation\, software\, hardware\, models\, and evaluation\
 , and explores why breadth matters when turning robot intelligence into
  something that actually works.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-break-may-17-may17-afternoon-break@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T072800Z
DTEND:20260517T074300Z
SUMMARY:Afternoon break
DESCRIPTION:Format: break\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may16-23@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T074300Z
DTEND:20260517T075300Z
SUMMARY:Prompts Don't Have Opinions. You Do.
DESCRIPTION:Everyone is talking about what AI can do. Nobody is talking 
 about what it leaves behind. Jay walks through where AI genuinely earns
  its place in his process: research\, documents\, quick prototypes. And
  where it consistently falls short: the interaction decisions\, the loc
 al context\, the innovation that doesn't exist in any training data yet
 . He gets into real examples from Canva Sheets\, Automations\, and othe
 r real life examples. The talk doesn't stop at craft. Jay gets honest a
 bout what design leadership actually costs\, how burnout here isn't tir
 edness but losing trust in your own eye\, and what it practically looks
  like to take care of yourself when your job is to be the thing AI cann
 ot replace.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-39@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T075600Z
DTEND:20260517T080600Z
SUMMARY:AI Needs Design Systems
DESCRIPTION:When AI can generate UI instantly\, the bottleneck is no lon
 ger creation. It is consistency. This talk explores how production comp
 onents\, design tokens\, and shared patterns help AI generate higher-fi
 delity prototypes that stay aligned with your product and are easier to
  ship.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-46@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T080900Z
DTEND:20260517T081900Z
SUMMARY:Should designers design? Balancing product vs polish
DESCRIPTION:The way designers and frontend engineers ship is changing at
  the literal speed of light. This talk looks at where the tooling lands
 cape stands today\, how end-to-end workflows are actually shifting\, an
 d what everybody can do to stay sharp no matter what’s in their design 
 toolkit.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-45@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T082200Z
DTEND:20260517T083200Z
SUMMARY:How to train and evaluate design intelligence across expressions
  of taste
DESCRIPTION:Unlike code\, design has no universal compiler. There is no 
 single “correct” output. It evolves across the creator\, the audience\,
  and the platform it lives in. This makes training\, evaluating\, and w
 orking with AI for design fundamentally harder than tasks with objectiv
 e ground truth.\nIn this talk\, we break down the core units of design 
 tasks and introduce ways to measure them through structured\, empirical
  benchmarks. We explore the kinds of data exposure required across cult
 ures\, contexts\, and expressions of taste\, and how to incorporate thi
 s to produce more faithful results.\nWe focus on iterative design as th
 e central paradigm and how to build systems where AI can be steered and
  refined across multiple steps in graphic design tasks\, while humans c
 an seamlessly take over for last-mile edits without resetting the desig
 n state.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-47@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T083500Z
DTEND:20260517T084500Z
SUMMARY:How to build a Company Brain
DESCRIPTION:Every company runs on know-how scattered across people's hea
 ds\, Slack threads\, docs\, and databases. Humans navigate that mess fi
 ne. AI agents can't. Connecting an agent to your CRM gives it access to
  your data\, but not understanding of your business\, which customer ma
 tters\, which thread is canonical\, what your team decided last week. T
 he fix is a Company Brain: a living map of how a business works that tu
 rns its artifacts into context any agent can use.\n\nFormat: talk\n\nSt
 atus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-41@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T084800Z
DTEND:20260517T085800Z
SUMMARY:Plan\, Preview\, Orchestrate: Three Primitives for Shipping with
  Cloud Coding Agents
DESCRIPTION:Cloud coding agents can write code all day\, but most teams 
 still struggle to ship with them reliably.\n\nIn this talk\, Heng Hong 
 shares lessons from building an AI-native product at YC\, and breaks do
 wn the three core primitives that will power every team’s cloud agents.
 \n\nPlan Mode: product design and visual MCQ decision-making\, where ag
 ents commit to a one-shot plan before they touch code.\nPreview Environ
 ments: full-stack sandboxes for every PR\, so agent output can be teste
 d the way a human would test it.\nCollaborative Task Orchestration: a K
 anban-style interface that lets builders coordinate many agents in para
 llel while keeping organizational context.\n\nYou’ll leave with a pract
 ical mental model for what cloud agents need around them — and why the 
 agent itself is actually the easy part.\n\nFormat: talk\n\nStatus: publ
 ic
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-40@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T090100Z
DTEND:20260517T091100Z
SUMMARY:All we're going to do is plan and review
DESCRIPTION:Coding agents do the coding. Humans do the planning and revi
 ew. The history\, why this trend will continue. How to actually set you
 r environment up for parallel agent execution\, and how to set yourself
  up to plan and review an obscene amount of AI generated code.\n\nForma
 t: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may16-24@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T091400Z
DTEND:20260517T092400Z
SUMMARY:The AI Model We Built for Deterministic Developer Tasks
DESCRIPTION:Most LLM systems today are built around a single transformer
 . In this talk\, I’ll present Interfaze\, an alternative architecture d
 esigned for deterministic developer tasks.\n\nInterfaze replaces the mo
 nolithic model with a hybrid architecture\, a stack of DNN/CNN-based pe
 rception modules for OCR\, charts\, PDFs\, and speech\, a perception la
 yer directly part of the model that transforms external data\, includin
 g web\, code\, and documents\, into a compact structured state\, and an
  execution layer for interacting with dynamic systems.\n\nThis shifts t
 he bulk of computation away from large models\, improving consistency a
 nd determinism when working with large unstructured data such as audio\
 , PDFs\, and tax documents\, where the margin for error is close to zer
 o\, while maintaining strong performance across benchmarks.\n\nFormat: 
 talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-44@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T092700Z
DTEND:20260517T093700Z
SUMMARY:No Country For Old Code: An open-source runtime for high-context
  agents
DESCRIPTION:Building agentic systems for the first-mile of data problems
  at scale doesn't give you the luxury of greenfield. We've had to meet 
 organizations and people where they are — to build reliable\, accountab
 le agentic systems that integrate into the middle of the stack instead 
 of the tip.\n\nFrom sentinels to declarative budgets\, here we'll cover
  the primitives and techniques we learned\, once we started from the as
 sumption that legacy isn't waste\, and that reliability comes from buil
 ding surfaces where knowledge can accumulate.\n\nFormat: talk\n\nStatus
 : public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may16-21@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T094000Z
DTEND:20260517T095000Z
SUMMARY:Give Agents Agency
DESCRIPTION:Most of the interesting work in AI engineering right now is 
 in the messy layer between a smart LLM and actually getting useful work
  done in the world. At Smithery\, we've lived that problem from the ins
 ide: bet on MCP\, went through the ups and downs of MCP vs CLI\, then w
 atched the harness era turn hand-rolled agent loops into standard infra
 structure. I'll discuss the lessons we've learned on the way\, and what
  we see coming next — emerging agent orchestration and how we might giv
 e agents more agency.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-may17-50@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T095300Z
DTEND:20260517T100300Z
SUMMARY:Optimizing Reliable Agentic Workflows: MCP\, Code Intelligence\,
  and Parallel Agents
DESCRIPTION:Agents are becoming active participants in how software is b
 uilt\, tested\, and operated. Making them useful on real projects requi
 res more than better prompts or larger models. It requires better workf
 lows: structural code intelligence\, MCP-native tools\, parallel coordi
 nation\, reproducible environments\, and fast feedback from real execut
 ion. This talk explores the primitives behind reliable agentic engineer
 ing workflows\, including code intelligence systems\, parallel coding a
 gents\, and tool surfaces designed for large codebases. The goal is to 
 show how developers can move beyond one-off prompting toward workflows 
 where agents have the context\, tools\, and feedback loops needed to co
 ntribute reliably.\n\nFormat: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
BEGIN:VEVENT
UID:talk-may-17-speaker-1778403240559@aie.65labs.org
DTSTAMP:20260517T163931Z
DTSTART:20260517T100600Z
DTEND:20260517T101300Z
SUMMARY:Closing from the AIE Team
DESCRIPTION:Format: talk\n\nStatus: public
LOCATION:The Capitol Kempinski Hotel Singapore
URL:https://ai.engineer/singapore#schedule
END:VEVENT
END:VCALENDAR