Inspiration — Opens with the "1 million drivers, spreadsheet dispatcher" hook, then goes into the specific pain points from the owner transcript: dead miles, guessed ETAs, mid-route hazards. Ends with the key insight that the data already exists in NavPro — it just needs an intelligent layer on top. What it does — Covers all 5 modules as clean bullet points, each starting with the gap name so judges can scan it quickly. How we built it — Technical depth without jargon: FastAPI + Claude tool-calling loop, agentic vs single-call distinction, React/Tailwind with the Trucker Path theme, and the mock data fallback strategy. Challenges — Honest and specific: the JWT + client-id auth issue you just debugged, the latency problem with sequential tool calls, and the FMCSA HOS rule complexity. Accomplishments — Leads with the agentic loop being genuinely reasoning (not rule-based), which is the strongest technical differentiator for judges. What we learned — The "talk to the expert first" lesson reads well as a hackathon insight judges appreciate. The explainability point ties back to the Marcus Webb example that's already in your demo. What's next — Six concrete directions, ordered from easiest (flip the mock flag to live) to most ambitious (fleet analytics).
Built With
- ai-/-ml-claude-claude-sonnet-4-6-(anthropic)
- anthropic-python-sdk-backend-python
- apscheduler
- axios
- fastapi
- geopy
- httpx
- lucide-react
- open-meteo-weather-api
- pydantic-frontend-react-18
- python-dotenv
- react-router-apis-&-integrations-trucker-path-navpro-api
- recharts
- resend-(email)-languages-python-3.11
- tailwind-css
- twilio-(sms)
- typescript
- uvicorn
Log in or sign up for Devpost to join the conversation.