AccidentAI

Inspiration

Car accidents are already one of the most stressful experiences in life. But the nightmare doesn't end at the scene—filing insurance claims takes weeks, requires endless paperwork, and often leads to disputes over damage assessments. We asked ourselves: what if AI could handle all of this in the 10 minutes you're still at the scene?

What it does

AccidentAI is an autonomous forensic agent that transforms your smartphone into a Level-3 Claims Adjuster. The app:

Guides evidence capture with real-time AI photo verification—telling you if images are too dark or blurry before you leave
Reconstructs crash physics using Gemini's multi-modal reasoning to calculate impact force, speed estimates, and collision modes
Generates forensic dossiers with ICAR-compliant damage assessments and repair cost estimates
Auto-files insurance claims via secure mTLS/OAuth2 handshakes with carrier APIs

How we built it

AI Core: Google Gemini 3 Pro for vision/reasoning and Gemini 2.5 Flash for real-time native audio assistance
Frontend: React 18 with Tailwind CSS and Lucide Icons
Security: Web Crypto API for forensic evidence hashing (cryptographic proof)
Architecture: Native Google GenAI SDK (zero-LangChain approach for ultra-low latency)

Challenges we ran into

Building a deterministic agent that doesn't just describe damage but actually reconstructs the physics was tough. We also had to ensure our evidence hashing system created legally-admissible cryptographic proofs.

Accomplishments that we're proud of

The live audio "Black Box" assistant that talks users through safety protocols in real-time, and the search grounding feature that pulls historical crash data from specific intersections.