A visualization of LLM sampling distributions at different temperature settings, showing probability mass shifting from peaked to flat.

Build a From-Scratch LLM Sampler: Temperature, Top-k/Top-p, and Multinomial Sampling vs. Hugging Face

A hands-on guide to building a from-scratch LLM sampler with temperature scaling, top-k/top-p warping, and multinomial sampling, benchmarked against Hugging Face. Includes runnable code and a roadmap to production.

September 20, 2026 · 14 min · 2888 words · martinuke0
A diagram showing a User Safety monitoring pipeline with event ingestion, processing, and alerting components.

User Safety: safe — A Hands-On Build Guide for Systems Engineers

A complete build guide for a User Safety monitoring system — ‘safe’ — that demonstrates distributed event processing, real-time alerting, and fault tolerance. Includes architecture diagrams, runnable Python code, and a senior-level extension roadmap.

September 20, 2026 · 11 min · 2321 words · martinuke0
A visualization of a Mixture-of-Experts architecture with sparse routing paths

Build a Sparse MoE Inference Engine from Scratch in NumPy

A hands-on guide to building a sparse MoE inference engine in pure NumPy, covering top-k gating, capacity-constrained token dispatch, and differentiable load balancing — a portfolio project that signals deep systems and ML engineering skill.

September 19, 2026 · 14 min · 2828 words · martinuke0
A visualization of ring attention with tokens flowing in a circular topology across GPU devices.

Build a Ring Attention Engine: A Hands-On Guide for Signal-Level Systems Engineers

A hands-on build guide for implementing ring attention — the technique behind near-infinite context transformers — as a portfolio project that signals deep distributed systems and ML engineering skill to hiring managers.

September 19, 2026 · 16 min · 3361 words · martinuke0
A Python code editor showing GGUF binary parsing logic and tensor dequantization

Building a Pure-Python GGUF Weight Loader with Dequantization for Llama Models

A hands-on guide to building a pure-Python GGUF weight loader with full dequantization support for Llama models — covering binary parsing, quantized tensor reconstruction, and a roadmap to production-grade infrastructure.

September 17, 2026 · 12 min · 2528 words · martinuke0
Feedback