Embedder2026 –
Engineer — YC S25
Hardware-aware AI for embedded software development. I build the agent itself — CLI, TUI, VS Code — and the hardware it touches: Infineon AURIX enablement, BLE and lab-instrument debugging.
2006/01/25
Leo Gao
I'm working on the AI–hardware interaction layer, where agents meet real machines.
Before that, I worked as an early engineer at Embedder (YC S25) and Wave RF — AI firmware agents, agentic tooling, and RF hardware.
I study Computer Engineering at the University of Michigan.
Work
Wave RF2025 –
Embedded Systems Engineer
Time-critical Bluetooth LE on Zephyr RTOS and Nordic nRF52 — real-time scheduling, sub-millisecond responsiveness, drivers adapted to Zephyr's device model.
SmartTO2024 – 25
Embedded Software Engineer
A $15k Smart Mobility grant to build C-V2X vehicular data oracles — Raspberry Pi 5 and C-V2X radios broadcasting live vehicle data in Toronto.
Selected Projects
Documents as a filesystem for AI agents — PDFs parsed into sections, tables, and figures an agent can grep, read in ranges, and cite by path and line. CLI + MCP server.
A CLI built for AI agents to work Reddit — browse, search, vote, comment, and batch-script against a shared client. Rich tables in a terminal, JSON envelopes when piped.
govcon-cli
Agent-first discovery of US (SAM.gov) and Canada (CanadaBuys) government contracts — domain presets for drones, robotics, and defense, one JSON contract, shipped as a Claude skill.
Upstream contributor — a merged fix to the AI SDK's Anthropic provider, allowing null content in the compaction-delta streaming schema.
Deterministic web animation to MP4 — drives a page's clock frame-by-frame through headless Chrome, with GPU capture 22× faster than the naïve pipeline.