I build production AI products — taking machine learning, deep learning, NLP and generative AI/LLMs from research all the way to systems that real people and businesses rely on every day.
Over the past 15+ years I’ve architected and deployed scalable, high-performance AI systems that serve hundreds of thousands of users. I care about the whole path: robust modeling, evaluation you can actually trust, and the data and MLOps engineering that keeps AI reliable in the real world.
I’m happiest working where research meets shipping — turning promising ideas into dependable products.
Products launched at Shoonya AI, Tealium, and Bayer / The Climate Corporation.
Real, revenue-generating AI products in production — not proofs of concept. A selection of the work that defines how I build.
Shoonya AI
A marketing technology platform combining generative AI content creation with behavioral intelligence — helping brands generate on-brand content at scale while measuring real business impact through LIFT measurement, turning campaign performance into clear, actionable insight.
Tealium
A real-time predictive audience and customer-data ML product delivering propensity insights on customer clickstream data at scale. I architected it end-to-end — from modeling to real-time serving and MLOps. $1.4M ARR · 100 customers.
Bayer / The Climate Corporation
One of digital agriculture’s leading farming platforms. I built the ML and data science that help farmers turn field data into decisions — including deep-learning yield forecasting and satellite field-mapping models powering decision tooling across one of agriculture’s most-used digital platforms.
Tealium
A generative-AI/LLM assistant built on a production RAG pipeline at Tealium, now serving 860+ customers. Together with AWS I built an evaluation framework using Ragas and Auto-Instruct on Amazon Bedrock, and balanced cost and latency with a mix of LLMs, function calling and fine-tuning. 860+ customers · ~85% context utilization · ~86% faithfulness · ~76% answer relevancy (AWS-measured).
How Tealium built a chatbot evaluation platform with Ragas and Auto-Instruct using AWS generative AI services.
AWS Machine Learning Blog · December 2024
A production-grade framework for evaluating LLM-powered chatbots — combining Ragas and Auto-Instruct on Amazon Bedrock to measure faithfulness, context utilization and answer relevancy at scale.
Work spanning marketing technology, agriculture, and energy — applying AI and quantitative modeling across each.
Rhode Island State Energy Plan — Business-As-Usual Energy Forecast.
Regional Greenhouse Gas Initiative (RGGI) Report 2016 — Part II.
EnergyVision 2030 — Technical Appendix.
Providence Climate Justice Plan.
Proof from building and shipping AI at scale.
Atharva AI Incorporated / Shoonya AI Pvt Ltd · Co-founder & Lead Data Scientist · 09/2026 – Present
Generative AI · LLMs · MLOps · Evaluation
Tealium Inc., Vancouver, BC · Staff Data Scientist · 06/2021 – September 2026
RAG · Fine-tuning · Function calling · Transformers · NER · GANs · Databricks · Seldon
The Climate Corporation (Bayer CropScience), St. Louis, MO · Sr. Data Scientist · 04/2019 – 05/2021
DNN / RNN · Python · CPLEX · Satellite imagery
Acadia Center, Hartford, CT · Policy Data Scientist · 09/2011 – 02/2019
Econometrics · XGBoost · CPLEX · Stochastic optimization
Duke University, Durham, NC · SENES Consultants, Delhi · Research Assistant / Jr. Environmental Engineer · Duke 09/2010 – 07/2011 · SENES 06/2008 – 07/2009
Operations research · CPLEX · Energy policy · Waste-to-energy
Honors: Nicholas School Merit-based Scholarship; Applied Statistics (R) Assistantship.
Building an AI product, exploring generative AI, or looking for a partner in applied ML research? Let’s talk.