At Fraynel Inc, I was the primary author and majority contributor to a Python LLM agent-orchestration platform that routes subtasks using LLM-as-judge decisions, with budget and permission limits enforced before execution. I also built a containerized evaluation harness that makes routing quality measurable on every commit.
I authored an MCP server and a multi-role agent runtime with GEPA prompt optimization gated by schema validation, fixture replay, and champion/challenger promotion. As a freelance Backend Engineer, Agent Infrastructure, I delivered a Python orchestrator for autonomous coding agents, isolating each node in its own Git worktree and branch.

