Transformers

Code editor showing a Python implementation of flash attention with tiling and online softmax in NumPy.

Building Flash Attention from Scratch in Pure NumPy: A CV-Worthy Systems Project

Build flash attention from scratch in pure NumPy using tiling and the online softmax trick, then extend it toward production. A deep, runnable project that signals real systems engineering to hiring managers.

Tiled matrix multiplication diagram with softmax rows

Building Flash Attention from Scratch in NumPy: A CV-Worthy Side Project

A working engineer’s guide to implementing Flash Attention from the ground up using tiling and the online softmax trick — runnable NumPy, end-to-end tests, and a roadmap to senior-level extensions.

Diagram of a mini-GPT training loop with mixed precision, gradient accumulation, and optimizer components.

Building a Mini-GPT From Scratch: A Portfolio Project That Signals Real ML Systems Skill

A working engineer’s guide to building a mini-GPT training loop from scratch in PyTorch — featuring AMP, gradient accumulation, cosine warmup, weight tying, and AdamW — designed as a CV-grade portfolio piece.

Two pipelines merging — a fast draft model feeding into a verifier that accepts or rejects proposed tokens.

Speculative Decoding: How Production LLM Systems Cut Latency by 2–3x

Speculative decoding trades a small draft model for big latency wins — 2–3x faster token generation with mathematically identical outputs. Here’s how it works and where production systems use it.

Diagram showing token positions feeding a transformer decoder with KV cache blocks reused across steps.

KV Caching in Production: How Transformers Trade Memory for Latency

A practitioner’s guide to KV caching: what gets cached, why decode steps are cheap, and how production systems like vLLM and TensorRT-LLM squeeze out more throughput.