Prabakaran Chandran

Prabakaran Chandran

MS Data Science @ Columbia · ML Engineering Lead · Causal Inference & Agentic AI

About Me

I am a graduate student at Columbia University pursuing a Master of Science in Data Science. My work spans the intersection of causal inference, machine learning, and artificial intelligence—with a deep commitment to understanding how agents, both human and artificial, make decisions under uncertainty.

Before Columbia, I led machine learning engineering at a stealth startup building agentic AI systems, and worked as a Senior ML Engineer at Informatica on enterprise AI solutions. My journey began at Mu Sigma, where I spent over three years as a Decision Scientist working across energy, manufacturing, pharma, retail, and smart cities.

I believe that the most meaningful problems lie at the convergence of human behavior and intelligent systems. My research interests include causal inference, probabilistic machine learning, reinforcement learning, and the design of AI systems that augment rather than replace human judgment.

What I Believe

Learning is Reconstruction

True understanding comes not from accumulation, but from the willingness to unlearn and rebuild from first principles. Like a diffusion process, we must sometimes embrace noise before clarity emerges.

Decisions Shape Destiny

Every meaningful outcome traces back to a decision made under uncertainty. Understanding the science of decisions—how agents choose, learn, and adapt—is understanding the substrate of progress itself.

Prosperity as Purpose

The ultimate measure of our work is whether it contributes to human flourishing. Technology, theory, and method are means; prosperity—broadly conceived—is the end that gives them meaning.

Patience and Craft

Great work emerges from sustained attention and care. There are no shortcuts to mastery. The willingness to sit with difficulty, to iterate, to refine—this is the price of work worth doing.

Areas of Focus

Data Science

  • Statistical Modeling & Inference
  • Causal Inference
  • Analytics Engineering
Read writings →

Machine Learning

  • Causal and Probabilistic ML
  • Reinforcement Learning
  • Theory of Interactive Decision Making
  • Modern Deep Learning
Read writings →

Artificial Intelligence

  • Continual Learning
  • Human-Centered AI
  • Foundation Models
  • Alignment, Safety & Interpretability
Read writings →

Current Work

Teaching at Columbia

Course Assistant · Causal Inference for Data Science Jan 2026 – Present

CA for Prof. Adam Kelleher's graduate course—a rigorous treatment of modern causal inference spanning potential outcomes, structural causal models, quasi-experimental methods, and ML for causal inference.

Teaching Assistant II · Advanced Analytics for Social Sciences (QMSS) Jan 2026 – Present

TA for advanced econometrics and applied causal inference bridging statistics and social science research.

Research & Experiments

Zero-Shot Coordination in Multi-Agent Systems — Formalized cross-play generalization as an objective in Dec-POMDPs. Built visuomotor policy architecture with residual CNN encoders, multi-head spatial attention, and GRU recurrence.

Hierarchical Causal State-Space Modeling — Developed a hierarchical causal recurrent SSM for retail basket time series enabling counterfactual simulation of marketing interventions.

Energy-Based Port-Hamiltonian Neural Networks — Designed differentiable physics engines that parameterize Hamiltonian energy for structure-preserving dynamics.

Education

Columbia University August 2025 – December 2026

Master of Science, Data Science

Analysis of Algorithms · Causal Inference · Probabilistic Machine Learning · Reinforcement Learning · Advanced Deep Learning & Generative AI

Anna University May 2015 – May 2019

Bachelor of Engineering, Control Systems · GPA: 8.63

Control Theory · Dynamical Systems · Signal Processing · Neural Networks & Evolutionary Algorithms

Explore

Get in Touch