PraCha
Prabakaran Chandran · பிரபாகரன் சந்திரன்
AI researcher · engineer · builder — Tamil Nadu → New York
I am interested in intelligence as both a technical and existential problem. That draws me toward , cognition, consciousness, philosophy, and metaphysics, but also toward building AI and ML systems that live at the intersection of frontier research, applied research, and real engineering. I believe more in learning than in static knowing, in deliberate experimentation over posture, and in interdisciplinary work as a method for discovering better questions and better systems. Each sprint of research is designed to produce a concrete result — working code, a finding, or a documented wall — and everything is published.
I bring 6.5+ years of experience across the stack of analytics and data science, with a career arc that has gone from sensors to tensors and has been shaped strongly by and complexity thinking. I am currently an MS Data Science student at Columbia, running FAIRE — a self-designed and self-directed research program — in parallel, working toward top AI labs and, in the longer run, toward building a decision-engineering company of my own. pracha.me is the visible trace of that process — building is the mechanism of becoming, and every shipped thing is evidence of who I’m turning into.
MS Data Science, Columbia University · Seeking Applied Scientist, Researcher & Research Engineer roles at frontier AI labs and enterprises · Available December 2026 · pc3197@columbia.edu · LinkedIn
Building Right Now
Oyster Club
Two people from different fields. One focused 100-minute session. The most interesting problem is rarely the one you arrived with — it appears when two worlds explain themselves to each other. A format built on the conviction that real thinking happens in building, not in decks: no agenda, no slides, just the problem that earns the room — and two curious people who won’t leave until something is made, shipped, or broken enough to learn from.
Join a session → AI/ML Studio · 125 days · New YorkManhattan Summer 2026
A self-directed AI/ML studio in New York. 125 days of deep work — one serious research paper, one product build, and everything made visible. The rule is simple: no shortcuts, no half-finished things, no ideas that never leave the notebook. Every sprint has a question at the start and a result at the end. The summer is the proof of method, not just output.
Explore themes → Research · -1 to 0Evolving Decision Systems
The right answer today is quietly the wrong one tomorrow — yet AI is frozen the moment it ships, and we cover the gap with armies of people rebuilding systems by hand. We’ve gotten good at building AI that’s smart once. The thesis: evolving decision systems from the intersection of AI, complex dynamical systems, and adaptive control — for clinical trials, health, life sciences, finance, and operations.
Explore the thesis →The Arc
- Systems & Complexity Control engineering taught me to see everything as a system — state, dynamics, feedback. Mu Sigma widened that to complexity: emergence, nonlinearity, how simple rules produce intricate behavior.
- Data Science Six years turning data into decisions — not reports that sit in decks, but models that organizations actually act on across industries.
- ML Engineering Satellite imagery, aquaculture, document understanding, enterprise AI. Systems that shipped, scaled, and changed how real operations ran.
- AI Research At Columbia: filling the theoretical gaps industry doesn’t have patience for. Causal reasoning, continual learning, reinforcement learning, probabilistic modeling.
- Decision Engineering The destination. Where everything converges — AI, causal thinking, complexity science — into systems that change how decisions get made at scale.
Selected Work
The Philosophical Roots of Frontier AI
A multi-part primer connecting philosophy, ML schools, lab strategies, reasoning systems, and the current frontier of AI research and engineering.
HMV-CRL: Separating What You Want from What You Were Shown
Reward functions conflate two things that shouldn’t be conflated: what the agent actually wants, and the observations it happened to see during training. This disentangles them — so the goal survives when the world changes.
Causal State-Space Models
Forecasting models predict what will happen — decisions require knowing what would happen. This integrates causal structure into time-series models so they can reason about interventions, not just patterns.
Zero-Shot Coordination in Multi-Agent RL
Agents trained separately fail to coordinate when paired with a stranger — they built different habits. This asks what training structure gives agents conventions general enough to work with anyone, without ever having met.
Latest Across The Site
- Loading…