Hyderabad, Telangana, India
I'm a software engineer with 3 years of experience building full-stack applications on the MERN stack. Over the last year I shifted into AI Engineering building GenAI systems that are reliable, controllable, and actually hold up in production My current work centers on Retrieval-Augmented Generation and the unglamorous parts that make it work: • RAG pipelines — chunking, embedding, and retrieval that stay grounded instead of hallucinating • Vector search — FAISS / embedding stores, tuning recall vs. latency, keeping context relevant • LLM orchestration — prompt structure, routing between models, tool-calling, controllable outputs • AI system design under real constraints — latency budgets, accuracy floors, cost, and offline/edge limits • Evaluation — measuring whether the system behaves, not just whether it responds I care less about what a model can do in a one-shot demo and more about how it behaves under real usage messy inputs, scale, and edge cases. Currently building GenAI systems and sharing the engineering decisions in public the trade-offs, the dead ends, and what actually shipped.
Designing and experimenting with GenAI systems including RAG pipelines, embeddings, and LLM-based services