About

Mehul Jain

Hi 👋, I’m Mehul

Data Scientist · Bayesian Modeling · Forecasting · NLP · Agentic AI

LinkedIn Github

Welcome to Thoughts, Code and Mischief — a corner of the internet where I write about applied modeling, machine learning, and the occasional detour into life.

The short version

I’m a Data Scientist who builds models that hold up under uncertainty and actually get used. My work sits at the intersection of applied Bayesian modeling, time series forecasting, NLP, and agentic AI — the kind of systems where a good estimate isn’t the finish line; it’s the start of a better decision.

How I got here

My fascination with data started when I realized that numbers aren’t abstract — they tell stories about systems, behavior, and what might happen next. That curiosity led me to Data Science at Northeastern University, where I learned to love the gap between a raw dataset and something that changes how a team operates.

Before grad school, I studied Computer Science at VIT, Vellore, and spent time on research and projects across ML, deep learning, and NLP. Those early experiments taught me something I still carry: the best technical work is useless if you can’t explain why it matters.

What I work on

Four areas show up again and again in my work:

  • Applied Bayesian modeling — hierarchical models, causal attribution, and posterior inference when you need uncertainty, not just a point estimate
  • Time series forecasting — demand, revenue, and operational forecasts that teams can plan against
  • NLP — turning unstructured text into structure: topic models, sentiment, retrieval, and document intelligence
  • Agentic AI — LLM systems that retrieve, reason, and take steps in a workflow, not just generate text

At Universal Music Group and Flynn Group, that has looked like Bayesian mix models to isolate what actually drives outcomes, revenue forecasting systems that replace guesswork with calibrated predictions, and NLP platforms that turn large text streams into signals people can act on. Alongside that, I build retrieval-augmented and agentic tools — including an in-house RAG system for data science teams — so models live inside real workflows.

The through-line: connect statistical rigor to the decision in front of someone, and ship it so it stays reliable.

Tools I reach for

PyMC / Bayesian inference Time series & forecasting PyTorch Scikit-learn NLP & LLMs RAG & agents Airflow MLflow Cloud pipelines

I’m most at home when a model moves from notebook experiment to something that runs reliably at scale.

What I’m looking for

I’m excited about Applied Scientist, Machine Learning Engineer, and Data Science roles where probabilistic modeling, forecasting, language systems, or agentic workflows are core to the problem — and where the output has to be both rigorous and usable.

Outside the code

When I’m not tuning models or wrangling pipelines, you’ll find me behind a drum kit or playing the Indian classical flute. Different kind of pattern recognition — but the rhythm helps.

Let’s connect

If you’re hiring for Bayesian modeling, forecasting, NLP, or applied AI systems — or you just want to talk shop — reach out on LinkedIn.