Sean Connell

I build software, embedded hardware, and live-broadcast systems for off-road racing — solo, end to end. Apps in the store, firmware in the field, video on air. If it has to work 200 miles from cell coverage, that's my kind of project.

One engineer, full stack — screen to silicon.

Most people pick a layer. I've shipped products at every one of these, usually wired together into a single working system.

Field-tested where failure is public.

Live race broadcasts don't get a second take. These systems ran for real audiences, in the desert, over Starlink.

// numbers from project docs & field logs — the receipts live in the repos

Five systems, one race weekend.

These aren't demos — they're an interlocking stack I designed, built, and operate for live off-road racing.

01

PreRun.io

Off-road GPS navigation app — design, build, ship, operate

Turn-by-turn navigation where roads don't exist. Import GPX routes, record trails by driving them, find new ones on satellite imagery, and navigate with audio guidance fully offline.

• Cross-platform Flutter app shipped to iOS, Android, web, macOS, and Linux
• Meshtastic mesh-radio integration for live group location with zero cell coverage
• Dual subscription rails — Stripe on Android/web, StoreKit 2 on iOS
• Self-hosted routing engine — a customized GraphHopper fork routing over OpenStreetMap merged with off-road trail data

FlutterFirebaseMapboxJava / GraphHopperMeshtastic

02

Autonomous GoPro Livestreamer

Firmware reverse-engineering & embedded Linux

A modified-firmware GoPro that becomes a self-healing livestream appliance: plug in power, it goes live over Starlink and stays live — recovering on its own from network outages, stream seals, and power loss.

• Reverse-engineered GoPro internals — custom C daemon and an LD_PRELOAD shim rewriting the camera's RTMP path
• Buildroot rootfs with a failure-recovery state machine, ported across multiple camera models
• Proven at the Baja 500 — six cameras ran a full race day unattended

CBuildroot / LinuxRTMPCloud Run

03

Race Broadcast Infrastructure

Live video pipeline — ingest, mix, transmit

The plumbing behind a live race broadcast: racers' phones and cameras push RTMP into my ingest layer; an operator dashboard pulls up to 50 feeds into a grid and cuts one program feed out through OBS to YouTube.

• HMAC-signed stream keys and private-network fanout over Tailscale
• Self-hosted on Hetzner — nginx-rtmp, ffmpeg, Caddy, systemd, with ops runbooks for race day
• Plus the on-air layer — OBS overlay graphics, data-driven leaderboards, and click-to-air racer bio cards

Node.jsnginx-rtmpffmpegBunny CDNOBS

04

Telebox — Vehicle Telemetry

Embedded hardware — CAN bus to live dashboard

A sub-$500 ESP32 box that plugs into a race vehicle's CAN bus, decodes ~30 telemetry channels at 10Hz across OBD-2, J1939, and MoTeC protocols, adds GPS + IMU, and streams live to a web dashboard over Starlink.

• Host-testable pure-C++ core with a clean HAL — 18/18 unit tests, CI, verified end-to-end against a simulator
• Pluggable per-vehicle decoders and offline buffering for connectivity gaps

C++ / PlatformIOESP32-S3CAN / TWAIFirebase RTDB

05

SatComms — Team Push-to-Talk

Real-time audio app — build to store-ready

A walkie-talkie app for teams spread across a course: low-latency push-to-talk audio with a shared live map, built to survive flaky connections with an auto-reconnect controller.

• LiveKit real-time audio with a Node token server and reconnect/backoff logic
• Full product build — subscriptions, cloud functions, store submission assets

FlutterLiveKitRevenueCatMapbox

The supporting cast.

Ship it. Run it. Stand next to it.

Have a problem that can't fail in public?

Let's talk.