Who am I?

About, briefly

I build machine-learning systems end to end and write down what I learn on the way. This is the running log.

Program
B.Tech · Data Science & AI
School
IIIT Naya Raipur
Years
2024 – 2028
Focus
Reinforcement learning · applied ML

I'm Shreyash Rai, a third-year B.Tech student in Data Science & Artificial Intelligence at IIIT Naya Raipur. I like taking a system apart to the first principles and rebuilding it myself — a physics simulator, a training loop, a retrieval pipeline — because that's the only way I actually understand it. What I learn doing that lands here as TILs; what I build lands as projects. If you want the short version of what I'm on right now, there's a /now page for that.

Education

B.Tech in Data Science and Artificial Intelligence at Dr. SPM International Institute of Information Technology, Naya Raipur (2024–2028), currently at a CGPA of 8.65/10. The coursework I keep coming back to: reinforcement learning, machine learning, linear algebra, probability & statistics, numerical methods, algorithms, and graph theory.

Research

Lagrange-Lock — reinforcement learning for satellite station-keeping, advised by Dr. Avantika Singh. I built a custom Gymnasium environment on a CR3BP physics core JIT-compiled with Numba (~10,000 steps/second), then trained a PPO agent over ~5 million timesteps with a reward balancing orbit retention, fuel use, and drift. It learned to hold Halo orbits for 5,000+ timesteps at the Earth–Moon L1 point; a Three.js viewer plays back its thrust trajectories.

FDTD prediction via DMD — numerical modeling, advised by Dr. Maifuz Ali. I modeled 1D FDTD electromagnetic wave propagation and applied Dynamic Mode Decomposition to predict field evolution, cutting computation roughly 10× versus direct simulation while holding accuracy.

What I build with

Mostly Python, with C++ and C when speed matters; PyTorch, Stable-Baselines3, and Gymnasium for the RL work; Transformers and sentence-transformers for NLP and RAG. The full list — languages, ML stack, scientific-computing and visualization tools — lives on the /uses page.

Elsewhere

The quickest read on what I'm up to is the projects page or the feed. To reach me, email is best.