Bookshelf.
A collection of items I've found to be valuable across the internet.
14 items
Articles
Build deep craft and range without outsourcing your judgment to AI ~ How to maximize your future in a rapidly changing world.
22580: From GPT-2 to Kimi K3, Explained
How LLMs evolved in storing and retrieving information ~ Great breakdown of LLM evolution.
How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance
A step-by-step CUDA matmul optimization worklog ~ One of the first things I read when learning CUDA.
Tensors, autograd, and training loops in one sitting ~ Solid intro to PyTorch fundamentals.
Inside vLLM: Anatomy of a High-Throughput LLM Inference System
Paged attention, continuous batching, and speculative decoding ~ vLLM deep dive.
TurboQuant: Redefining AI Efficiency with Extreme Compression
Extreme compression for KV caches and embeddings ~ Revolutionary quantization with zero accuracy loss.
Videos
Using AI as a personal tutor ~ The inspiration for a learning harness I'm building.
Stanford CS229: Machine Learning (Autumn 2018)
Andrew Ng on the foundations of ML ~ The "Gold Standard" for ML courses.
Large Language Models explained briefly
A short animated intro to LLMs ~ Quick visual grounding for what happens inside one.
Transformers, the tech behind LLMs
Transformers and attention, visualized ~ Awesome look at the core architecture of an LLM.
Papers
FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning
Better warp partitioning for faster attention ~ Significant improvement on FlashAttention.
LoRA: Low-Rank Adaptation of Large Language Models
Fine-tune LLMs with low-rank weight updates ~ Incredibly efficient fine-tuning.
Repositories
GPU fundamentals to kernel optimization to production inference ~ Still working through it, really great stuff.
Apple's ML framework for Apple silicon ~ Been messing around with it on my Mac (ported models coming soon).