CS and physics undergraduate (biology minor) at CSULB, BUILD Scholar, aiming for an MD/PhD. I build scientific software and agentic AI systems, and I care most about measurement: tools whose results you can inspect, reproduce, and check.
Research: behavior and neurobiology · protein molecular dynamics · gait and rehabilitation · collaborative surgical-ergonomics work. Independent work: human–AI collaboration and agent systems — building them, studying where they fail, and finding when a procedure can be reused reliably instead of reasoned through again.
| Project | What it is |
|---|---|
| workflow-checks | Evidence-first back-office checks: freight detention and carrier invoice audit, PT authorization worklist, utility bill exceptions. Python stdlib, tested in CI, synthetic data. Live sample outputs → |
| CaptureSuite | Windows-first multimodal research capture: daemon, versioned worker plugins, sealed sessions, PySide6 operator UI. |
| CanvasPilot | Agentic Canvas LMS tools for LLMs — an MCP server + CLI driven by your own authenticated browser session. |
| TowerOps | Synthetic air-traffic decision-support demo (Strands Agents) where deterministic checks gate every state change behind conflict projection, approval, and readback. |
| uma-sim | Unofficial Umamusume career simulator in Rust. |
Hackathon builds: testpilot (agent that writes, runs and repairs regression tests for a diff) · ledgerly (invoice + payout copilot with a human approval gate) · tastetable (weekly outing planner for seniors and caregivers on Qloo taste signals).
Each repository documents its scope, evidence, and limitations. Research contributions are collaborative, not claims of independent clinical validation.
Open to research collaborations and scoped pilots: understand an existing process, build a bounded alternative, and measure quality, time, cost, and human intervention against it.
Outside the lab: cosplay, physical making, music, anime, games.