Pallav Mahapatra
Open to workJoining Taskrabbit as Senior Data Scientist · April 2026
I started my career building data pipelines — the unglamorous foundation that everything else sits on. Somewhere in those early years at Amazon I developed an obsession with a specific question: when a metric moves, how do you actually know if you caused it?
That question pulled me out of BI engineering and into data science, through a part-time master's degree, and eventually into building causal inference systems and GenAI evaluation pipelines at Amazon scale. The core problem never changed — it just got more interesting. How do you measure something rigorously enough to bet a product decision on it?
This site is where I write down what I've figured out. Not tutorials — more like the things I wish someone had told me before I spent a month going down the wrong path.
Where I've worked
Senior Data Scientist
IncomingTaskrabbit
Data Scientist
Amazon · Customer Experience
Built GenAI evaluation infrastructure for shopping assistants before launch — LLM-as-Judge frameworks, multimodal pipelines, and the human audit systems needed to trust the outputs. Also ran large-scale A/B experiments measuring causal lift across 450K products.
Data Scientist
Amazon · Seller Experience
Built a distributed causal inference platform in PySpark to estimate treatment effects at GB scale. Used DoubleML, DR Learners, synthetic control, and WGAN-based counterfactual generation — mostly to answer the question "did this account manager intervention actually do anything?"
Lead BI Engineer
Amazon
Owned the metric layer for a $10B lead management platform. 50+ analysts and data scientists relied on it daily. This is where I learned that the hardest part of data work is not the model — it's getting the numbers right before you touch statistics.
Education
MSc, Data Science
Liverpool John Moores University, UK
B.E., Electronics & Telecommunication
University of Mumbai, India
Let's talk
Always happy to connect.
Whether you're thinking through a causal inference problem, building experimentation culture at your company, or just want to talk about something you read here — reach out.
Writing
Thoughts on pieces I've published, questions, pushback — I read everything.
Collaboration
Research, guest posts, or technical deep-dives with people working on similar problems.
Opportunities
Interesting problems in measurement, causality, or AI evaluation at well-run teams.