About

Backend & AI engineer building tools for agent runtimes, CI, and observability. AgentDiff is my current focus trajectory regression testing that diffs AI agent runs as DAGs and gates CI on drift, loops, and cost. It's shipped to PyPI, has a live website, and caught a real tool-loop regression on a live Gemini agent Before this I built high-throughput financial engines under compliance (SEBI), orchestrated legal knowledge graphs with Neo4j + Gemini, and was a founding full-stack AI engineer at Omara Technologies. I contribute to LiteLLM and DeepEval in my spare time and run an automated CS newspaper, The Latent Chronicle. I work in Python, Go, and TypeScript happy to chat if you're building agents or need a hand on tough backend/AI problems.

Badges

Tastemaker
Tastemaker
Gone streaking
Gone streaking

Maker History

  • AgentDiff
    AgentDiffTrajectory regression testing for AI agents
    Aug 2026
  • 🎉
    Joined Product HuntJuly 10th, 2023

Forums

•

2mo ago

AgentDiff - Trajectory regression testing for AI agents

Traditional evals test outputs, missing when agents loop on tools or burn 50% more tokens. AgentDiff fixes this by recording agent runs as DAGs and diffing execution trajectories against golden baselines directly in CI. It catches tool loops, cost spikes, and latency regressions before merge evaluating how the model arrived at the answer, not just what it returned. Framework-agnostic with adapters for OpenAI Agents, Langfuse, LangSmith, and OpenInference.
View more