Prabakaran Chandran

I’m Prabakaran Chandran (Pracha)—an ML Engineer / Data Scientist with a long-running interest in interdisciplinary science and research, especially where ideas have to survive real systems: uncertainty, feedback loops, incentives, and human constraints. I’m currently pursuing an MS in Data Science at Columbia University (2025–2027), and I support graduate courses as a Teaching Assistant in Causal Inference and Advanced Analytics.

Before Columbia, I spent 6.5+ years building and shipping ML/DS systems across industry—document understanding workflows, enterprise AI tooling, forecasting and time-series modeling, computer vision, optimization, and simulation. The work has been varied, but the pattern has stayed the same: I’m most interested when the problem isn’t just “predict well,” but decide well—when the system changes because you acted, and the data changes because the world responded.

The anchor I keep returning to is:

Systems × Behavior × Intelligence.

Systems: dynamics, constraints, stability, emergence—how things evolve over time.

Behavior: people, markets, organizations—how decisions actually get made.

Intelligence: learning, inference, representation, planning—how capability is built.

That’s also how I think about AI. To me, AI is not an end in itself; it’s a capability layer inside larger human systems. The interesting questions are rarely “can a model do X?” but: What happens when the model is deployed into a living system? What incentives does it create? What feedback loops does it trigger? What fails silently? What remains reliable under distribution shift? I’m drawn to approaches that are evidence-driven, interpretable when it matters, and robust under real-world messiness—and to a view of AI that’s fundamentally about co-existence: humans, tools, and models shaping outcomes together.

Through PrachaLabs, I’m organizing my learning, writing, and experiments around a long-term theme I call Agentic Decision Sciences—a practical lens for studying decision-making across humans and machines in shared environments. Lately I’ve been spending most of my time around dynamical systems and time series, causal thinking and invariance, modern deep learning, reinforcement learning (including multi-agent settings), continual learning and memory, and interpretability through simulation and counterfactual reasoning—mostly as notes, prototypes, and benchmarks that compound over time.

Outside the screen, I’m a book-and-notes person. I like bookstores, long cafe sessions, sketching diagrams, and reading across complexity/system dynamics, behavioral economics/cognition, and the foundations of learning and inference—then translating what I learn into clear write-ups and small experiments. I try to keep my pace steady: no FOMO, no hurry, no comparison—just consistent work.

Contact: pc3197@columbia.edu · pracha.me