AI / COMPUTING · λ 590 NM · PROJECT
SafePulse
Real-time PPE compliance detection. First place, Infineon × YETI hackathon.
Problem
BGHM — the German trade association for wood and metal — records 123,707 workplace accidents, 10,283 serious injuries and 381 fatalities in its sector. The dominant root cause isn't equipment failure. It's protocol negligence: protective equipment not worn, or worn incorrectly, in the moment it mattered.
Safety is also an ESG metric almost nobody instruments. Emissions and energy get measured continuously and reported to investors; worker safety gets counted after the fact, in incidents.
Approach
SafePulse detects PPE compliance from the existing camera feed in real time, alerts the worker and the supervisor at the moment of non-compliance, and gamifies the response so the correction becomes habit rather than reprimand.
The third part is the one most detection systems skip. A system that only flags violations becomes surveillance, and shop floors defeat surveillance. Framing compliance as a score the worker owns changes who the system is for.
Result
First place, 24-hour Infineon × YETI hackathon.
I led a team of five and split us into two parallel workstreams — AI prototype and business case — so that neither the demo nor the pitch was assembled in the last hour. We shipped a working MVP and the winning pitch inside the window.
Team: Anton, Sanchit, Shankar, Prakrisht, Pushpit.
No repo. Demo screenshots from the deck not yet supplied.