Utsav Mishra
Lead Data Scientist with 7+ years building production machine learning, reinforcement learning, and GenAI systems at scale, including a patent-pending RL recommendation engine at Target.
Lead Data Scientist
About Me
I’m a Lead Data Scientist focused on production ML, RL, and GenAI—turning research into systems that hold up at scale (recommendation, personalization, forecasting, and experimentation).
As Lead Data Scientist at Target, I lead a team of 18 DS/AI engineers for Roundel AdTech: real-time sponsored-ads relevance, LLM-powered labeling pipelines, and GenAI/RAG initiatives with clear business metrics. Earlier, as a Senior Data Scientist, I led offer personalization and invented CORE, an RL-based contextual bandit with a filed patent.
Outside of work, I like long walks, treks, travel, and music.
Professional Focus
- Technical leadership: ML strategy, experimentation design, and production architecture across cross-functional teams.
- Research to production: End-to-end ownership from modeling choices to deployment and monitoring; CORE delivered major redemption/opt-in uplifts and $100M+ in incremental sales annually.
- People: Mentorship, leveling frameworks, and reusable ML patterns that raise the bar org-wide.
Key Expertise Areas
- Reinforcement Learning: Multi-armed bandits, contextual bandits, DDQN, exploration-exploitation strategies
- Generative AI: RAG-based retrieval systems, LLM integration, automated ML workflows
- Deep Learning: LSTM, CNN, neural architectures, transfer learning
- ML Infrastructure: GCP, VertexAI, Kubeflow, PySpark, TensorFlow, MLOps pipelines
- Causal ML & Experimentation: A/B testing frameworks, causal inference, statistical validation
Academic background
B.Arch (Honours) from Indian Institute of Technology Kharagpur. Coursework included programming and data structures, probability, image processing and remote sensing, and economics. Hackathons and quantitative projects included placements at the Microsoft Azure Hackathon and MishMash Hackathon.
Work Experience
Lead Data Scientist, Target Corporation
May 2024 — Present
• Developed a real-time query-item relevance model to filter irrelevant sponsored ads, classifying pairs (exact, substitute, complement, irrelevant) using E5 embeddings and a deep learning model; achieved a 9.2% relevance lift at 10 ms p99 latency.
• Designed a Gemini-powered LLM labeling pipeline to generate training data for the relevance classifier, categorizing 15M+ query-product pairs with attribute extraction and human-in-the-loop audits.
• Led TargetRun, an LSTM-based customer-level trip-count forecasting system that reduced MAPE by 50% and improved coupon opt-in and redemption across customer engagement segments.
• Architected Lookalike, a GenAI and retrieval-augmented generation (RAG) system for campaign forecasting, achieving 100% coverage at K=3 and eliminating up to 8 manual hours per campaign.
• Lead 18 DS/AI engineers for Roundel AdTech; mentored 5 team members, 3 of whom were promoted this year.
Senior Data Scientist, Target Corporation
October 2021 — May 2024
• Invented and filed a patent (U.S. App. 63/573,806) for CORE, a reinforcement learning-based contextual bandit for dynamic offer recommendation, driving 70% higher redemption and 28% higher opt-in (Mar–Nov 2023) and generating $100M+ in incremental sales annually.
• Owned the ML strategy and technical roadmap for Target's offer personalization platform (100M+ users), defining experimentation frameworks, model architectures, and scaling patterns.
• Standardized cross-team ML training, pipelines, and monitoring using GCP, Vertex AI, Kubeflow, and Grafana.
• Received the Data Sciences VP Award for spearheading RL optimization for Target Circle.
Applied Data Scientist 2, Tata AIA Life
April 2020 — October 2021
• Applied NLP techniques including Gensim LDA topic modeling to identify drop-off points in customer onboarding, achieving 6.2% conversion rate increase. Developed interactive journey visualizations using Plotly Sankey diagrams.
• Developed a GBM regression profit prediction model (R² = 0.991) and deployed it in production via FastAPI; reduced valuation TAT by 30 days. Implemented SHAP for feature attribution and built Streamlit visualization for explainable AI.
Machine Learning Engineer, HDFC Life
June 2019 — March 2020
• Developed a Face Recognition system with Anti-Spoofing capabilities using CNN and DeepFace embeddings to reduce fraud in virtual customer onboarding. Deployed as a containerized service using Docker and ECS.
• Designed and developed the serverless architecture of Lead allocation and Route optimisation app on AWS Data Lake (Lambda, DMS, Dynamodb, Step Function, S3, API Gateway, and Docker).
• Built a reinforcement learning-based insurance underwriting risk classifier, achieving 91%+ accuracy with only 40 questions.
Education
Indian Institute of Technology, Kharagpur
Bachelor of Architecture (Honours), conferred August 2019
July 2014 — April 2019
Relevant Coursework : Programming & Data Structures, Probability - Mathematics 1 & 2, Image Processing & Remote Sensing, Economics Micro & Macro.
Extracurricular : Sponsorship & Marketing Head, Spring Fest; Website Secretary, S'SAP Council; Lead National Service Scheme Unit - 5.
Recipient of the Merit-cum-Means Scholarship (2015–2018), awarded by the Ministry of Human Resource Development, Govt. of India.
Kendriya Vidyalaya, Andrews Ganj
Science PCM C++
April 2010 — April 2013
Percentage : 88.7% in XII (AIISCE).
Cum GPA : 9.8 in X (AISSE-CBSE)
Received Merit Certificate from CBSE for prolific academic performance in AISSE &AISSCE in 2011 & 2013.
Portfolio