Building a durable execution runtime in Rust for nondeterministic AI agents. It journals every step of a run so agents survive any crash, resume exactly where they stopped, and replay faithfully for debugging and audit.
Backed by Agentical, a probabilistic process calculus for durable multi-agent LLM systems, with Lean 4 proofs and a Rust reference runtime.
Supported hardware QA for home security systems by configuring and integrating devices (locks, sensors, lighting) into automated testing workflows. Performed hands-on lab work including soldering buck converters and embedded boards, assembling RJ45 Ethernet cables, and provisioning Raspberry Pi units for test automation. Built an internal software tool for lab device management.