Hi there 👋

Hey, I'm Nandeshwar Gupta.

I currently work as Manager, Machine Learning Engineering with Zupee.

What drives me?

If you ask me on any random day, I will say it's Music and Travelling. But nowadays I am neck deep in Generative AI & Reinforcement Learning.

Experience

Manager, Machine Learning Engineering Apr’25 - Present

• Led a 5–7 member ML team to build a RL-based recommendation engine, evolving from Multi-Armed Bandits (MAB) to Deep Q-Networks (DQN), driving +3.5% revenue and +10 BPS retention on a 2M+ DAU platform.
• Developed user embeddings from behavioral & demographic clustering, boosting recommendation accuracy and personalization engagement.
• Built an LLM-driven agentic chatbot with orchestrator agents, tool-calling, and context/nudge flows, improving CSAT by 8%.
• Launched scalable LLM chatbot with RAG, vector DBs, async pollers, and queue scaling, enabling 100% MoM growth and a 4-star Play Store rating.
• Delivered Learning Platform, utilising RAG + Elasticsearch for generative search & chatbot.
• Designed LLM framework with observability, routing, and prompt management for production.
• Built SQL Smith, a graph-based agent for NL-to-SQL, enabling faster self-service analytics.

Lead Machine Learning Engineer Jan’24 – Mar’25

• Built and deployed real-time tournament recommendation with Apache Flink + Aerospike, cutting inference from hours to seconds and driving 4.15% revenue growth in partnership with product and CRM teams.
• Improved recommendation accuracy@top4 by 6% and redesigned architecture to solve cold start issues, accelerating business impact realization.
• Developed multi-armed bandit pricing optimization system, delivering 8% revenue uplift.
• Parallelized model updates with Airflow, reducing training cycles from hours to minutes.
• Drove Generative AI projects including avatar generation, gender/NSFW detection, and celebrity voice cloning for IVR campaigns with marketing agencies (+6% call completion, +4% reactivation).
• Ran supervised cross-functional campaigns with CRM/marketing, reducing customer acquisition cost and boosting reactivation rates.

Senior Machine Learning Engineer Sept’22 – Dec’23

• Spearheaded end-to-end ML lifecycle framework and real-time prediction system (training, tracking, deployment, monitoring) from scratch.
• Delivered tournament recommendation engine boosting revenue by 4% and improving accuracy@4 by 10%, enhancing user experience.
• Led Generative AI PoCs: LLM-powered quiz question generation, Indian-dialect TTS for game commentary, marketing video generation, and SageMaker GPU training optimization.

Senior Data Scientist, V-Mart Retail, Feb2021–Present

Machine Learning

• Built Recommendation Engine for V-Mart E-commerce Platform.
• Customer Segmentation Analysis to increase sales.
• Customer Churn Prediction using Light GBM.
• Covid-19 Forecasting using Time-Series Analysis.

Data Engineering

• Built ML Pipelines using tools like Airflow, Celery & MLflow on AWS & on-prem for scheduling jobs.
• Collaboratively developed Data Lakehouse Architecture using AWS S3, Glue & Athena.
• Built CI/CD pipelines using AWS pipelines & GitHub actions.

Manager Data Science, Vasitum, Aug 2020 – Feb 2021

Machine Learning

• Built Recommendation Engine from scratch to drive better engagement & conversions.
• Chatbot – Built & maintain NLP based chatbot on RASA framework with Entity detection & Intent Classification.
• Elasticsearch – Optimized search engine to provide relevant results & improve entity discovery experience.
• Built end-to-end machine learning solutions (NLP, NER, DNN) to solve problems such as Spam Job classification, etc.

Data Engineering

• Database modeling for large-scale data collection.
• Transition to serverless architecture using AWS Lambda & Fargate

Data Scientist, Quick Company, May 2018 – August 2020

Machine Learning

• Applied Named Entity Recognition (NER) on a scale using Spacy.
• Implemented a Text Classification model on legal text using WordEmbeddings.
• Built Captcha Recognition model using CNN & OpenCV.
• Optimized Image Classification model built on PHOG algorithm.

Data Engineering

• Developed a fully automated event driven scraping framework.
• Used Neo4j database to link data & create graph structure.
• Built dashboards for reporting using D3js & Plotly.

Backend Services

• Created a unified search engine using Elastic Search.
• Built various applications for Web Backend & APIs using Django, DRF, boto3 & automated tasks (scrapers, schedulers, queue-based jobs) on AWS.
• Dockerized various APIs (microservices)

Tools

1. Packages - Pytorch, Tensorflow, Langchain, Langgraph, Streamlit, Gradio
2. Services - AWS, Athena, S3, Git, Bitbucket, Jenkins,
3. Languages - Python, SQL, Linux, Bash, C++