At Agent Orchestrator, I built core orchestration for parallel AI coding agents in isolated Git worktrees, with PR tracking and automated CI and code review feedback loops. I fixed a broken ETag / HTTP 304 guard, batched reads into GraphQL, and added caching to reduce API calls by 92%.
I also built durable mid-session switching between Claude Code and Codex, token usage and cost tracking, and a CDC-driven notification system. At Shipd by Datacurve, I converted ambiguous coding and debugging requirements into executable RL evaluation environments, hidden tests, and scoring rubrics for LLM agents.
