AI systems from agents to robots.
I build and evaluate autonomous software, Physical AI, robotics simulation, and the engineering infrastructure that makes complex AI behavior measurable and inspectable.
A preregistered Isaac Sim and ROS 2 Franka campaign: 20 seeds × 2 conditions, 40 scheduled and valid runs, with raw evidence and paired analysis preserved. Filtering improved the three headline paired metrics in 20/20 seeds, while 0 of 40 runs passed the full acceptance gauntlet. Thresholds were frozen before execution and were not relaxed afterward. The result is simulation-only: no physical-hardware, safety, certification, production-readiness, or real-world-transfer claim.
Read the commit-pinned M4 case study or watch the short case study.
A bounded agentic workflow that parses systems-engineering requirements, generates and executes tests, and returns traceable evidence for each check.
A publication-oriented research export covering physics-informed AI, robotics simulation, evaluation, MBSE, edge deployment, manifests, provenance, and checksums.
Franka Panda pick-and-place in NVIDIA Isaac Sim using Lula inverse kinematics, PhysX, and ROS 2 integration.
A robotics glossary and SysML v2 model library for interfaces, decomposition, and systems traceability.
Deterministic repository preflight and release-drift checks.
Across agents and robots, I care about evaluation, harness and tool design, provenance, reproducibility, negative evidence, requirements traceability, and technical judgment that survives inspection.