Work that gets clearer when you open it.

Reliability fixes and developer-experience improvements across production-scale open-source ecosystems.

• Reproduce platform edge cases and reduce them to reviewable, testable failures.
• Work through public feedback and adapt changes to the maintainers’ existing contracts.
• Improve reliability, documentation, CI behavior, and contributor workflows without widening the patch unnecessarily.

Agent memory, multi-agent coordination, verification, and safe execution systems built from first principles.

• Design memory lifecycles that balance recall quality, provenance, latency, and context cost.
• Build orchestration graphs with explicit state transitions, tool boundaries, retries, and human checkpoints.
• Create evaluation harnesses that turn model behavior into repeatable engineering evidence.