SourchesAI a multi-modal AI research engine processing 10,000+ daily documents, cutting research workflows by 75% and saving users an average of 6 hours per research project through automated data extraction. Stack: React, Python, FastAPI, Groq, Tavily, Upstash, Firebase, Vercel, Render • Built robust Python/FastAPI data pipelines integrating Groq, Tavily, and Upstash, and utilized embedding models with Pinecone vector databases to synthesize and distill noisy inputs into high-fidelity, structured knowledge bases with strict, automated source attribution. • Developed evaluation frameworks using PyTorch to measure and improve model performance across key generative AI metrics.