Headline

Senior AI Engineer — production RAG systems and LLM agents

About

I build the backend for enterprise generative AI products — currently three systems, each end-to-end on Azure: a legal helpdesk that answers questions over uploaded contracts and drafts legal paperwork, a patent assistant with semantic search over a large patent corpus, and a capital-expenditure analytics platform. The stack is Python, FastAPI, Azure OpenAI (GPT-4o), Azure AI Search, and LangGraph. Most of my work is the unglamorous part of making LLM systems dependable: production RAG with layout-aware document parsing and hybrid retrieval with semantic reranking; schema-validated structured outputs; agents that recover from failed tool calls; and grounded-refusal guardrails, so the system declines rather than invents. An evaluation harness scores groundedness on a fixed test set and gates every prompt and retriever change before it ships — evaluation is part of the system, not an afterthought. Outside work I build open-source developer tooling around LLMs and the Model Context Protocol (MCP), and I contribute upstream to the LLM ecosystem. Before this, I worked in applied data science: Bayesian marketing-mix modelling in Google Meridian at Analytic Edge (C5i), NLP and causal inference on pharma problems at Axtria, and an internship at AbbVie building predictive models on longitudinal patient data. That grounding in statistics and classical machine learning shapes how I build with LLMs: measure first, then ship. MS in Business Analytics from UC Irvine, where I was a Dean's Scholar.

Location

Hyderabad

AI stack

anthropic apis

antigravity

claude code

claude

elevenlabs

gemini

Work experience

Senior AI Engineer at Motherson Technology Services, Nov 2025 - Present · 10 mos, Hyderabad, Telangana, India, • Build the complete backend for three enterprise generative AI products on Azure. Stack: Python, FastAPI, Azure OpenAI (GPT-4o), Azure AI Search, LangGraph, Azure Document Intelligence, Docker. • Legal Lens, an AI legal helpdesk: production RAG over uploaded documents — Azure Document Intelligence parsing, layout-aware chunking, hybrid search with semantic reranking. Answers questions, compares contracts, and drafts legal paperwork as schema-validated JSON with grounded-refusal guardrails. • LangGraph tool-calling agent that plans and runs multi-step drafting and document workflows, invoking retrieval, comparison, and generation tools and recovering from failed calls. • LLM evaluation harness scoring groundedness and faithfulness on a fixed test set; gates every prompt and retriever change. Request tracing and token/cost/latency monitoring in production. • IP/patent assistant: RAG semantic search over a large patent corpus, supporting the full patent-filing lifecycle. • Capital-expenditure platform backend (in progress): ingests data from an upstream AI system and uses GPT-4o structured outputs to generate insights on interactive dashboards.

Data Science Consultant at Analytic Edge, May 2025 - Oct 2025 · 6 mos, Bengaluru, Karnataka, India, • Designed a Bayesian full-funnel marketing-mix model in Google Meridian: media impact across awareness, consideration, and conversion, with sequential dependencies to isolate incremental per-channel attribution. • Built a Streamlit app for interactive budget reallocation at national, regional, and sub-regional levels, with automated re-runs of impact simulations.

Data Scientist & AI Consultant (Freelance) at Freelancer.com, Oct 2024 - May 2025 · 8 mos, Bengaluru, Karnataka, India, • Built an AI document Q&A chatbot for client document collections using LangChain and RAG over Google Gemini. • Built a stock-market analysis agent combining real-time market data (yfinance), news sentiment (NewsAPI), and LLM-generated insights, delivered as a Streamlit app with Altair visualisations.

Data Science Intern at AbbVie, Jan 2024 - Jun 2024 · 6 mos, Irvine, California, United States, • Built predictive models on longitudinal disease and medication data to guide patient-care strategies. • Trained LSTM embeddings of medical concepts; 86% accuracy in disease-progression prediction.

Data Science Associate at Axtria - Ingenious Insights, Apr 2022 - Jun 2023 · 1 yr 3 mos, Gurugram, Haryana, India, • Engineered an NLP pipeline extracting medical entities from social media; domain-adapted PubMed word embeddings via incremental training on Twitter data; distance-based extractive summarisation for healthcare insights. • Measured the incremental impact of a pharmaceutical partnership program using propensity-score matching and hypothesis testing.

Analyst at Axtria - Ingenious Insights, Jul 2021 - Apr 2022 · 10 mos, India

Education

MS, Business Analytics, University of California, Irvine - The Paul Merage School of Business

BE Manufacturing Engineering + MSc Biological Sciences (dual degree), Birla Institute of Technology and Science, Pilani

Public links

github: https://github.com/lets-order-some-fries