ML leader with experience building production ML systems and academic research in applied mathematics and scientific machine learning. I enjoy developing and deploying models with mathematical rigor for technological and scientific applications.

Leadership

At The New York Times, I lead a team of ML scientists developing and deploying real-time ML models for subscription-related problems, such as serving a paywall at optimal moments or personalizing user-facing messages. I drive strategy, execution, and alignment of model development with business outcomes. Our production models serve millions of users and generate tens of millions of dollars in revenue, using causal ML, contextual bandits and personalization.

Industry (Individual Contributor)

Previously as an IC and tech lead at The Times, I developed and deployed real-time causal ML models for multi-objective optimization of our paywall (see blog post). I have also built large-scale production batch models that determined user-level access to our content (see blog post, PyData talk, VentureBeat article). I wrote statically-typed production code in Python and Go for high-throughput inference on cloud-based containerized infrastructure.

At the Times, I have also applied causal inference to observational data using statistical matching and sequential trial emulation, and designed Randomized Control Trials for training causal ML models. At Amazon, I built a Double (De-Biased) ML model to estimate the causal impact of Advertising products.

Research (PhD & Masters, MIT)

My Ph.D. thesis advised by Prof. Jörn Dunkel (Applied Mathematics), combined scientific ML, numerical computation, and fluid dynamics. I built a computational inference framework to learn Partial Differential Equations (PDEs) from microscopic data of active matter (see PNAS article). I also contributed to early demonstrations of Neural Differential Equations in learning physical phenomena (see preprint). I was a recipient of the Mathworks Engineering Fellowship.

I also received a Master’s degree from MIT for my work on theoretical and experimental modeling of the fluid dynamics of internal waves. My B.Tech in Mechanical Engineering at IIT Madras involved theoretical fluid dynamics research and building an omni-directional robot for ABU Robocon (that’s me driving!).

Outside of work, I enjoy long-distance running in the summer and skiing in the winter.

See my CV for more details.

Skills

Machine Learning

Supervised and unsupervised machine learning, Contextual bandits, Bayesian methods (Thompson Sampling), Causal Inference, Causal Machine Learning, Neural Differential Equations, Scientific Machine Learning

Mathematical Modeling

Differential Equations, (Fluid) Dynamical Systems, Optimization, Linear Algebra, Numerical Methods for PDEs

Technologies

Programming

Machine Learning

ML Deployment