https://fahim.bd Notes on machine learning, statistics, neural architecture search, and AI for development. en https://fahim.bd/blog/scaling-laws-for-llms https://fahim.bd/blog/scaling-laws-for-llms Mon, 04 May 2026 00:00:00 GMT The most expensive equations in AI determine how labs spend billions. Here's what they actually say — and where they're being rewritten. From Kaplan to Chinchilla to inference-time scaling. https://fahim.bd/blog/llm-quantization-demystified https://fahim.bd/blog/llm-quantization-demystified Tue, 28 Apr 2026 00:00:00 GMT Your 7B model has 14 billion numbers. Here's exactly how to shrink them — and what you lose in the process. A practitioner's guide to choosing GGUF, GPTQ, or AWQ. https://fahim.bd/blog/mixture-of-experts-explained https://fahim.bd/blog/mixture-of-experts-explained Mon, 20 Apr 2026 00:00:00 GMT How DeepSeek-R1, GPT-5, Gemini, and Mistral Large 3 all use the same trick — and what it means for your work. A complete conceptual and technical guide to MoE architecture. https://fahim.bd/blog/causal-inference-ml https://fahim.bd/blog/causal-inference-ml Sun, 15 Mar 2026 00:00:00 GMT Most ML models in global development optimize for prediction accuracy. I argue this is the wrong objective — and what we should be doing instead. https://fahim.bd/blog/nas-edge-deployment https://fahim.bd/blog/nas-edge-deployment Tue, 20 Jan 2026 00:00:00 GMT Building ML models that work on low-power edge hardware in climate-constrained settings taught me more about model design than any benchmark ever did.