# ML Systems

> Deep dives on machine learning systems: inference, training infrastructure, GPU memory, quantization, and LLM architecture — articles, primers, and case studies, with interactive tools alongside the writing.

Machine learning, from kernels to clusters. Written for engineers who want the mechanics, not just the API surface: how attention and KV caches actually work, how LoRA fine-tuning changes weights, how GPU memory gets consumed during training and inference, how transformer masking works, and how to reason about throughput and quantization trade-offs.

## Blog
- [Blog index](https://mlsystems.dev/blog/): All articles.
- [How transformers work — attention and KV cache](https://mlsystems.dev/blog/how-transformers-work-with-attention-and-kv-cache/)
- [Neural networks from zero](https://mlsystems.dev/blog/neural-networks-from-zero/)
- [LoRA fine-tuning guide](https://mlsystems.dev/blog/lora-fine-tuning-guide/)
- [Masks in transformers](https://mlsystems.dev/blog/masks-in-transformers/)

## Interactive tools
- [Playground index](https://mlsystems.dev/playground/): Interactive tools for building intuition.
- GPU memory calculator, throughput calculator, kernel benchmark, tokenizer explorer, attention visualizer, attention pattern atlas, eval harness, model card generator.

## Topics
- [Topics index](https://mlsystems.dev/topics/): Articles grouped by subject area (inference, training infrastructure, GPU memory, quantization, LLM architecture).

## Full index
- [Sitemap](https://mlsystems.dev/sitemap-index.xml): Every page on the site with last-modified dates.