Talks
Ideas I've shared on stage
Stop Asking Your Agent To Behave
📍 AI Engineer Singapore, Singapore
🗓️ May 17, 2026
- Why allowing agents to skip permissions isn't secure permission prompts are not scalable
- Introduced Fence as the policy and enforcement layer for execution boundaries to contain agent overreach
- Swiss cheese model for secure agent execution (classification, policy, isolation)
Designing AI to Scale Human Thought
📍 AI Engineer World's Fair, San Francisco
🗓️ Jun 3, 2025
- Why we should build AI for human augmentation in a presently automation-based landscape
- UX patterns (blind spot detection, cognitive partnerships, proactive guidance)
- Design principles (trust framework, progressive enhancement)
Concepts for Reliability of LLMs in Production
📍 Google Developer Workspace, Singapore
🗓️ Oct 11, 2023
- Concerns with utilizing LLM APIs in production ML systems
- Swiss cheese strategy for reliability in LLM-enabled applications
- Compartmentalization, explainability, online guardrails, offline testing, human evaluation
- Examples / case studies