title: DriveCore
emoji: 🚗
colorFrom: indigo
colorTo: purple
sdk: docker
pinned: false
DriveCore 🚗⚡
An AI-powered incident response, forensic analysis, and branch debugging platform for autonomous vehicle fleets.
Built for the AMD Developer Cloud Hackathon — Track 1: AI Agents & Agentic Workflows Powered by Qwen3 running on AMD Instinct GPU via LangChain multi-agent pipeline.
Quick Start Instructions to Run on AMD GPU Droplet:
Start Ollama ollama serve &
Start Qwen backend cd /app/Autopulse uvicorn backend:app --host 0.0.0.0 --port 8006 &
Start frontend bun run dev --host 0.0.0.0 --port 30000 &
Go to Website http://165.245.137.74:30000
What is DriveCore?
DriveCore routes every autonomous vehicle incident, near miss, and sensor log through a multi-step AI agent pipeline — surfacing root causes, compliance concerns, and operator coaching plans in seconds.
🤖 AI Agent Pipeline
Step 1: Intake Agent → Classifies incident, extracts vehicle ID, timestamp, severity Step 2: Enrichment Agent → Identifies affected AV subsystems Step 3: Risk Agent → Root cause analysis ranked by confidence Step 4: Response Agent → Immediate actions, short-term fixes, long-term prevention Step 5: Documentation Agent → Formal Markdown safety report + Qwen's Personal Recommendation
🏗 Tech Stack
- AI Model: Qwen3 via Ollama
- Agent Framework: LangChain
- GPU Compute: AMD Instinct MI300X via AMD Developer Cloud
- Backend: FastAPI (Python)
- Frontend: React 19 + TanStack Start + Vite + Tailwind CSS v4
🏆 Hackathon
Built for AMD Developer Cloud Hackathon — Track 1: AI Agents & Agentic Workflows
Powered by AMD Instinct GPU ⚡