Vaibhav Nakrani

ML and Agentic AI, explained without the hand-waving.

einsum, one index at a time

Every PyTorch codebase I read eventually hits a line like this:

Vaibhav Nakrani

NVFP4 Is a Scaling System, Not Just a 4-Bit Number

What the payload, block scale, and tensor scale each do, and when a 4-bit claim holds up

Suppose a model tensor is stored in NVFP4. It is tempting to picture every entry as a miniature, self-contained FP16 value: four bits go in, one approximate real number…

Vaibhav Nakrani

A Visual Guide to Causal, Bidirectional, and Hybrid Inference in Video Generation

How information flow changes across video time and denoising time

“Causal” and “bidirectional” sound like labels attached permanently to a video model. One model moves left to right; another sees the whole clip. Pick an architecture…

Vaibhav Nakrani

Diffusion Without the Hand-Waving

The cleanest way to understand diffusion is to build it up one step at a time. Here’s the version that finally made it click.

Vaibhav Nakrani