Silicon · firmware · software · web
Schematic, layout, bring-up, firmware, the network it lives on, and the page you watch it from. Nothing is lost in translation between four vendors, because there aren't four vendors. I bring my own tooling and a library of proven hardware blocks, so the first prototype arrives in weeks — not after a quarter of scaffolding.
A board that cannot talk is a paperweight. I design the link with the circuit — radio budget, topology, segmentation and security decided before layout, not bolted on after the first field failure.
Proprietary toolchain and a library of universal hardware blocks — carrier boards, driver stacks, a bring-up rig — already written and already proven. Your budget buys the part that is actually yours, not the scaffolding around it.
Models wired in through your own API keys — no reseller markup, no data leaving on my account. Vision that catches the defect a tired eye misses, plain-language search across a decade of service records, anomaly detection on live telemetry.
My default for instruments and motion. G0 and G4 for cost-down control loops, F4/F7 when there's a display or a filesystem in the way.
Kinetis and LPC for long-lifecycle medical and industrial builds where a ten-year supply guarantee is worth more than a cheaper part.
Where the thing needs to talk. Wi-Fi and BLE provisioning, MQTT telemetry, OTA update channels with a way back.
Architecture, part selection, second sources named up front.
Stack-up, impedance, thermal, EMC done at layout — not after.
Rails, clocks, first blink, and a test fixture built alongside.
Drivers, state machines, bootloader, OTA with a way back.
API, database, auth, the telemetry pipeline behind the device.
The dashboard, the app, the page your customer actually sees.
The site you are reading was designed and built by me — the layout, the motion, the video backgrounds and the code. The GitHub contribution graph above is real data on a scheduled job, not a screenshot. That is the capability on offer: your site can show live data — machine status, order pipeline, inventory, sensor feeds, whatever your business actually runs on.
Medical, lab and defense software is not ordinary software with extra paperwork — the paperwork is the engineering discipline. Requirements traced to tests, changes traced to reasons, and a build you can reproduce two years from now. Retrofitting it later costs about five times what doing it from the first commit does.
| Standard | What it means in the repo | Role |
|---|---|---|
| IEC 62304 | Software safety classification, unit-level requirements, SOUP inventory for every dependency you did not write | LEAD |
| 21 CFR Part 11 | Electronic signatures, immutable audit trails, record retention that survives a database migration | LEAD |
| ISO 14971 | Hazard analysis that reaches the firmware — every failure mode with a mitigation you can point at in code | LEAD |
| IEC 60601-1 | Programmable electrical medical systems — the firmware clauses, watchdogs and safe states | ADVISE |
| Cybersecurity | Signed firmware, secure boot, SBOM, and a documented path for patching a device already in the field | ADVISE |
Your schematic, layout or codebase read properly. You get the risk list, the part-availability problems, and what I would do differently.
Board, firmware, service, interface — any slice or all of it. Delivered working, documented, with the sources.
Firmware releases, part obsolescence, the web app, and the 2 a.m. question about why a board in the field went quiet.