Python

Hands-On: Scratch‑Built GPT Training Loop with Activation Recomputation in Pure Python

A minimal, pure‑Python GPT training loop with gradient checkpointing, perfect for showcasing systems‑level deep learning know‑how on a CV.

Diagram of a paged KV cache with request preemption

Building a Continuous Batching LLM Serving Engine with Paged KV Cache and Request Preemption

Learn how to design and implement a high‑throughput LLM serving backend that batches requests continuously, uses a paged KV cache for memory efficiency, and preempts low‑priority requests to maximize throughput.

Cover image depicting a vector index architecture

Building a Pure-Python IVF-PQ Vector Index with On-Disk Inverted Files

A hands-on guide to implementing an IVF-PQ vector index in pure Python, with on-disk inverted files and product quantization for scalable similarity search.

A diagram of a transformer model with highlighted KV cache

Building a Pure‑Python Int4 KV‑Cache for Transformer Inference

This project demonstrates low‑level memory optimization, quantization, and efficient inference techniques that signal strong systems engineering expertise to hiring managers.

Diagram of a hierarchical sliding window context manager

Hierarchical Sliding Window Context Manager with Semantic Eviction and Priority-Based KV Cache

A practical guide to implementing a hierarchical sliding‑window context manager with semantic eviction and priority‑based KV cache in Python, perfect for showcasing systems expertise.