Exploring the mathematics of intelligence
I'm Gaurav Bhowmick — a student of quantitative methods, artificial intelligence, and machine learning. This is where I document what I'm learning, building, and questioning.
About
This site is my public notebook: a place to explore ideas, visualize concepts, and share what I'm figuring out. I think in distributions, not averages.
I'm drawn to the elegance of optimal transport theory, the surprising depth of attention mechanisms, and the practical beauty of well-calibrated models.
I'm not here as an expert — I'm here as a student who believes the best way to learn is to build and share.
Click a node to explore
This graph maps the topics I'm studying and how they connect. Each node is an area of curiosity, and the edges represent how these ideas flow into each other. Nothing exists in isolation — that's what makes this fun.
Foundation
Fell down the probability rabbit hole
Started seriously studying measure-theoretic probability and stochastic calculus. Built my first Monte Carlo simulations and realized how beautiful randomness can be when you give it structure.
Deep Dive
Discovered Optimal Transport
Read Villani's "Topics in Optimal Transportation" and was hooked. Started implementing Wasserstein distance computations and Sinkhorn algorithms. Applied clustering methods to time series regime detection.
Architecture
Deconstructing transformers, layer by layer
Spent months understanding attention mechanisms from first principles. Built visualizations of how LLMs process multimodal inputs — from image patches to token embeddings. Created an explainer video series breaking down the inference pipeline.
Generative
Exploring diffusion models and flow matching
Studied the mathematical connections between score-based diffusion, denoising, and flow matching. Implemented toy diffusion models to build intuition for how noise becomes signal.
Current
Building my public learning lab
Launched this site as a home for interactive explorations. Currently studying reinforcement learning from human feedback, information geometry, and the intersection of optimal transport with generative models. @Marginator on X for data-driven takes with edge.
Interactive Lab
Neural Network
InteractiveGradient Descent
Click to placeDistribution Morphing
Optimal TransportAttention Heatmap
TransformerBrownian Motion
StochasticExplorations
Wasserstein Distance: An Intuitive Visual Guide
Building intuition for earth mover's distance through interactive visualizations and worked examples.
How Does an LLM Actually Process an Image?
A visual walkthrough of multimodal transformers — from pixel patches to token embeddings.
Regime Detection with Clustering: A Learning Notebook
Exploring how unsupervised methods can identify hidden states in time series data.
Let's explore together
I'm always looking for interesting conversations about AI, quantitative methods, and the ideas that sit between disciplines.