Abhay Sharma

At RadialBrain, I tuned a RAG pipeline to reach 80% answer relevance on a 50+ question test set and reduced unsupported answers by 40%. I indexed 900 documents with FAISS and exposed LLM-powered features through async FastAPI services.

At XEX, I shipped 10+ REST APIs and cut p95 query latency by 45% across five endpoints by redesigning PostgreSQL schemas, adding indexes, and rewriting queries. I also secured production endpoints with JWT authentication and rate limiting.

I built FactLayer, a cross-document intelligence engine that extracts structured facts from enterprise PDFs with bounding-box grounding and citation verification. Its pipeline uses Gemini 2.0 Flash, FAISS, and a Neo4j knowledge graph to reduce unsupported claims by 40%.

My projects also include research on adversarial attacks and certified robustness for DreamerV3 world models, and a ConvNeXt–Swin ensemble for skin lesion detection that reached 94% accuracy. I’m completing a B.Tech in Computer Science and Engineering at Vellore Institute of Technology (VIT-AP University), and I’m seeking AI/ML, Machine Learning, or Generative AI Engineer roles worldwide.