Exploring AI and High-Performance Computing | Generated by AI
Question: What does this blog contain? (https://www.aleksagordic.com/blog)
Answer:
The blog at https://www.aleksagordic.com/blog is maintained by Aleksa Gordić, a research engineer with experience at DeepMind and Microsoft, focusing on AI, machine learning, and high-performance computing. Its tagline is “Computers can be understood,” and it aims to make complex computing concepts accessible through in-depth technical explanations.
Main Themes
- GPU architecture and low-level optimizations — Especially NVIDIA GPUs, matrix multiplication (matmul) kernels, PTX/SASS assembly, warp tiling, and tensor cores.
- High-throughput AI inference systems — Such as vLLM, including paged attention, continuous batching, prefix caching, and large language model (LLM) serving.
- Machine learning fundamentals and learning guides — Reinforcement learning (RL), geometric/graph ML, transformers, NLP, and general ML starting points.
- Historical and philosophical AI topics — Breakdowns of foundational papers (e.g., Alan Turing’s “Computing Machinery and Intelligence”) and speculative pieces on post-artificial superintelligence (ASI) worlds.
- Personal experiences — Career advice (e.g., landing jobs at DeepMind), learning strategies, and journeys into AI.
The posts are highly technical, often featuring code analysis, diagrams, and deep dives into state-of-the-art systems, while earlier posts include educational guides for beginners in various ML subfields.
Selection of Recent and Notable Posts
- Inside NVIDIA GPUs: Anatomy of high performance matmul kernels (September 2025) — Deep dive into GPU matmul optimizations.
- Inside vLLM: Anatomy of a High-Throughput LLM Inference System (August 2025) — Detailed explanation of vLLM’s architecture and techniques.
- Neocambria (June 2025) — Philosophical discussion on humans in a post-ASI world.
- ELI5: FlashAttention (2023) — Simplified explanation of FlashAttention.
- Landing a job at DeepMind (2021) — Advice on securing a research engineer role without an ML degree.
- Learn RL / Learn Geometric ML / Learn transformers / Learn ML (2019–2021) — Guides for getting started in these areas.
- Computing machinery and intelligence - Alan Turing (2020) — Two-part breakdown of Turing’s seminal paper.
- Learning how to learn (2019) — Tips from a Coursera course on effective learning.
The blog appears to have been revived in 2025 with advanced technical content after a hiatus, building on earlier educational posts.
References: