I build production AI systems end to end: LLM applications, agents, retrieval, backends and the infrastructure around them. I studied CSE (AI & ML) at VIT-AP, graduated in 2025, and have been shipping at Bettrlabs since October 2025.
I work backwards from the use case. I map the workflow, design the architecture and decide how data moves through the system, and only then decide where a model fits. Most of what I have built began as a systems-design problem; the prompt came last.
At Bettrlabs that has meant a legal and regulatory RAG, a compliance system for FMCG products, hybrid search over formulations, Razorpay payment infrastructure and a large part of the platform backend. Retrieval is one tool I use, alongside agents, tool orchestration, state and plain backend engineering.
I take full responsibility for the systems I deliver: the architecture, the system logic and the debugging decisions, and I make sure I can explain every part of what goes to production. When something breaks I work from logs and traces of what the system actually did, and I watch tokens, latency and cost as closely as correctness. Outside work I build Jarvis, a personal assistant on an agent harness I wrote, to learn each layer of an agent by hand.