Headline

Sr. Data Scientist at UHG | 3 patents | IIM - A

About

I build AI systems that turn messy, unstructured documents into decisions worth millions. Since 2018, I’ve worked at the intersection of machine learning, NLP, and computer vision — and for the last five years I’ve led AI initiatives at UnitedHealth Group (Optum), one of the world’s largest healthcare companies. My team’s work spans LLM-powered RAG pipelines, document intelligence, knowledge graphs, and semantic search: systems that map policy information from complex contracts to databases and save an estimated $35M by eliminating 4–5 hours of manual work per policy across 500K policies. Along the way, I’ve filed three patents (DigiSense, LayoutSense, TableSense), earned UHG’s Inventor Award, and shipped models that process claims with 98.7% accuracy. What makes my profile a little different: while leading a full-time ML product team, I completed an MBA at IIM Ahmedabad (Class of 2026). That means I don’t just speak model architecture — I speak P&L, pricing strategy, corporate finance, experimentation and causal inference, negotiation, and operations strategy. I’ve learned to evaluate an AI project the way a CFO would, and to explain one the way a product leader should. Where I add the most value: • Leading data science and ML engineering teams from problem framing to production (Azure, Kubernetes, MLOps) • Building GenAI/LLM products that survive contact with real enterprise data • Translating between technical teams and business stakeholders — roadmaps, ROI cases, and executive communication I’m exploring senior data science, ML leadership, AI product, and strategy roles — remote-first, hybrid in Gurgaon/Bangalore, or international. If you’re building something ambitious with AI, let’s talk: jatin_lamba@icloud.com

Location

Bengaluru

AI stack

anthropic apis

codex

langchain

langgraph

langsmith

openai

Work experience

Senior Data Scientist at UnitedHealth Group, Apr 2025 - Present · 1 yr 5 mos, Gurugram, Haryana, India, Agentic Enrolment Forms Automation • Built an end-to-end agentic pipeline that takes enrolment forms from inbox to system of record with no human touch — email ingestion and attachment parsing, LLM-based form extraction, then agentic classification and routing. • Designed a multi-agent orchestration layer (LangChain / LangGraph) where extraction, validation, classification and correction agents hand off to each other, with self-correction loops and confidence thresholds deciding when a human is actually needed. • Deployed LLM-based structured extraction over GPT-4o and Azure Document Intelligence, using prompt chaining, few-shot templates and schema-constrained JSON output to handle 100+ form layouts without writing per-layout rules. • Straight-through processing on roughly 70% of incoming forms, with average handling time down from 25 minutes to under 5. • Added evaluation harnesses and drift monitoring so extraction accuracy is tracked per form type rather than as one aggregate number. OGA Finance — End-to-End Document Automation • Owned the full automation pipeline for Optum's finance document workflows: ingestion, OCR, table and line-item extraction, reconciliation logic, exception handling and downstream system posting. • Combined layout-aware document AI with LLM reasoning to pull structured financial entities from invoices, statements and remittance documents that had no consistent template. • Removed close to 200 hours of manual processing a month across the workflow and cut downstream rework by around 40%. • Shipped it properly: containerised services on AKS, CI/CD through GitHub Actions, MLflow tracking, API endpoints consumed by internal UIs. Filed 3 patents — DigiSense, LayoutSense, TableSense — and received UHG's Inventor Award. Python, PyTorch, TensorFlow, LLMs (GPT-4o, Gemini), LangChain, LangGraph, RAG, Agentic AI, Azure AI Search, Azure Document Intelligence, FAISS, MongoDB, FastAPI, Docker, Kubernetes, MLOps.

Data Scientist at UnitedHealth Group, Aug 2022 - Mar 2025 · 2 yrs 8 mos, Gurugram, Haryana, India, Formulary Digitization • Built the digitization pipeline for formulary documents — thousands of heterogeneous PDFs and scans turned into structured, queryable drug and tier data. • Developed a document segmentation model combining OpenCV features with NLP signals, hitting ~90% accuracy classifying page objects (tables, headers, footnotes, multi-column blocks) before extraction ever ran. • Built table digitization that converts nested formulary tables into a knowledge graph, improving data accessibility by 82% so downstream teams query relationships instead of parsing rows. • Used Siamese networks and RNN-based sequence models for template matching and layout similarity across inconsistent document families. Contract Intelligence • Built semantic search over the contract corpus using embeddings, approximate nearest neighbour retrieval and clustering, replacing keyword lookup with meaning-based retrieval and improving search efficiency by 90%. • Designed and deployed ML models to extract KPIs and key clauses from PBM contracts, taking 2-4 hours of manual review off each document. • Trained text classification models to categorize clauses and paragraphs for downstream analytics and automation. • Later extended into a RAG pipeline (Azure AI Search + MongoDB) with API endpoints powering an internal UI — mapped policy information from contracts to databases at ~85% accuracy, saving 4-5 hours per policy across 500K policies, roughly $35M in avoided effort. Provider Contract Automation • Optimized automation workflows in COSMOS, bringing manual data loading from ~100 minutes down to 15-20 and delivering $1M+ in savings across two phases. • Deployed LLM-based entity extraction with prompt tuning to identify contractual entities from unstructured text where rules-based parsing kept failing. • Took models from notebook to production personally — Torchserve, Docker, Azure, monitoring — rather than handing off at the model artifact.

Associate Data Scientist at UnitedHealth Group, May 2021 - Aug 2022 · 1 yr 4 mos, Gurugram, Haryana, India

Assistant Manager - Business Analyst at Altruist Technologies Pvt. Ltd., Sep 2019 - Apr 2021 · 1 yr 8 mos, Greater Bengaluru Area, - Implemented business intelligence solutions and conducted advanced analytics to identify opportunities and strengths to meet business goals. - Proposed and implemented outbound customer service strategies with customer risk scoring. - Identified inefficiencies at the associate level and introduced KPIs to help improve productivity and performance.

Machine Learning Engineer at KoiReader - Digitize Unstructured Data, Nov 2018 - Sep 2019 · 11 mos, Bangalore, Worked with the development team in the development of a product that: - Uses NLP, Computer Vision, and newest advancements in technology to extract information from pdfs. - Extracts information from text and image and all kinds of pdf. - Gives the information output in a very readable and handy format.

Management Trainee at Indiabulls Real Estate, Jul 2017 - Aug 2017 · 2 mos, Gurugram, Haryana, India, • Accumulated data on Real Estate outline of Haryana andMaharashtra. • Developed visualizations on the changing Real Estate structure. • Produced an Impact Analysis report of GST, RERA and demonetization. • Curated a survey to analyze the sentiment among buyers.

Education

Postgraduate Degree, Masters of Business Administration, Indian Institute of Management Ahmedabad

Post graduate program, Business Analytics, Praxis Business School