Downstream Damage Predictor
An AI-powered automotive diagnosis platform that analyzes vehicle failures, predicts downstream damage, and generates structured repair recommendations using LLMs.
The Narrative
Built an AI automotive diagnosis system using FastAPI, Next.js, Ollama, and Gemini to analyze vehicle issues and generate structured repair procedures, replacement parts, possible cascading failures, and preventative maintenance recommendations.
Implemented a modular backend architecture with Pydantic validation, multi-provider LLM support, and dependency injection for scalable AI service management.
Developed persistent semantic caching using ChromaDB to store and retrieve previous diagnosis results, reducing redundant LLM inference and improving response latency.
Integrated structured JSON generation and validation pipelines to ensure reliable AI responses that can be directly consumed by the frontend application.
Tags
Links
System Highlights
LLM Diagnosis
Uses Ollama and Gemini models to generate vehicle-specific failure analysis and repair recommendations.
AI Response Cache
Persistent ChromaDB-based semantic caching avoids repeated model inference for similar diagnosis requests.
FastAPI Backend
Modular API architecture with Pydantic validation, service layers, and provider-based LLM integration.
Next.js Interface
Modern frontend for vehicle input, model selection, and structured diagnosis visualization.
