30-Day Edge AI MVP Sprint

I’m kicking off an intense 30-day, around-the-clock push to prove a rugged, offline-first edge-computing stack that ingests live video streams and RF sensor feeds, runs lightweight Edge AI models to detect and classify objects, fuses those detections into coherent tracks and alerts, and presents everything in a single operational picture through a simple web UI.

Edge AI skill is the keystone, so experience with on-device inference (TensorRT, ONNX, OpenVINO, etc.) and squeezing models onto constrained hardware is essential. A background in robotics, real-time systems, SDR, or computer vision will feel right at home in this sprint.

The MVP must visibly demonstrate:
• Continuous live video ingestion from IP/USB cameras
• Real-time fusion of RF sensor data with video detections
• A lightweight browser UI that renders the operational picture as clear graphical charts/plots (map layers or dashboards are nice extras)

Everything runs untethered at the edge, must survive intermittent connectivity, and recover after power loss. ATAK integration is a welcome bonus but not required.

Success means I can boot the system on ruggedised hardware, watch video and RF feeds roll in, see fused tracks refresh in <1 s, and receive alerts when thresholds trip. I’ll need reproducible build scripts (Docker-compose or k8s), concise documentation, and a short screen-capture walkthrough of the working demo.

We’ll move fast: daily stand-ups, GitHub issues, rapid code reviews, CI/CD with self-hosted runners. Go, Rust, or C++ back-end plus a TypeScript/React front-end is my preference, but I’m flexible if you can hit the performance bar.

Tell me which slice of the stack you want to own and how you’ll lead it to the finish line in thirty days.