Details

Adopt AI

Senior Forward Deployed AI Engineer

I build AI automation infrastructure and work directly with customers on integrations: API discovery from HAR recordings, coding-agent workspaces, secure remote execution, and authenticated browser sessions.

Live 3D architecture

Use the navigator to move between diagrams and related work, then scroll for more details.

Delivered API classification and input/output schema inference from repeated HAR calls. This work later became coding-agent subagents in ABCD.

My work

Forward-deployed AI work that connects product integration, execution policy, and customer workflows.

I shipped the API schema inference layer that classifies repeated HAR calls, uses specialist agents to vote on fixed versus optional inputs and outputs, and keeps integrations repairable through fresh recordings when private web APIs change.
I designed and built ABCD from scratch, alone: a Python toolkit and coding-agent workspace for discovering APIs, developing workflows, testing against the platform and publishing actions.
I forked and re-architected Alibaba Open Sandbox for Adopt AI's RCE layer, owning the platform integration, WDL fallback integration, security posture, nonblocking command lifecycle, sandbox TTL keep-alive, pod-to-pod routing, and warm capacity work.
I delivered anonymized forward-deployed customer workflows, including inventory forecasting with ML plus LLM recommendations and human approval, and handwritten document automation that turns mixed scans into structured spreadsheet and filing artifacts.
I helped design and develop Tabby with teammates, providing browser sessions for human login handoff, authenticated automation and browser-based execution.

Stack and domains

  • Python
  • TypeScript
  • Kubernetes
  • AWS EC2
  • Karpenter
  • Temporal
  • Open Sandbox
  • Docker
  • WDL
  • jq
  • Cursor
  • Claude Code
  • LLM orchestration
  • HAR analysis