I graduated from IIT Guwahati in 2020 with a B.Tech in Mechanical Engineering and a Minor in Product Design. I got interested in computer science early in college and never looked back.
For past 6 years I have built all kind of AI/ML systems, starting from recommender systems and computer vision all the way to NLP, document AI and LLM agents these days. I have worked at a large company like Tata Digital and at small startups, across e-commerce, insurance, legal tech and conversational AI.
What I am working on now
I am building Shrota, an AI phone-survey platform, on my own. Apart from that I run my own research on structured outputs: can an LLM write only the part of a schema-bound output that changed, and splice it back with the schema still valid? In my experiments on Gemini, edits took 172x fewer output tokens. The code is on GitHub.
I am also studying speech processing. I wrote up what I learned about text-to-speech in a post you can listen to.
Skills
- LLM systems: RAG, multi-agent systems, tool calling, structured outputs, LangGraph, LangChain, MCP
- Evaluation: Langfuse, Ragas, OpenTelemetry, guardrails
- Voice: LiveKit, speech-to-text, text-to-speech, turn detection
- ML: PyTorch, Hugging Face, scikit-learn, spaCy, OpenCV
- Data and search: Milvus, FAISS, Neo4j, PostgreSQL, Redis
- Backend and infra: Python, FastAPI, Docker, Kubernetes, AWS, GCP, GitHub Actions
Courses
Fine-tuning and RL for LLMs, efficiently serving LLMs (vLLM, SGLang), agentic design patterns, knowledge graphs for RAG, and the 3D Vision Summer School at IIIT Bangalore in 2024.
Want to work together? Let's talk