Sam Austin AI https://samaustinai.pages.dev/ A Machine Learning & Artificial Intelligence Blog en Docker Compose for ML Projects: Multi-Container Development Setup https://samaustinai.pages.dev/2026/09/docker-compose-for-ml-projects-multi https://samaustinai.pages.dev/2026/09/docker-compose-for-ml-projects-multi Wed, 30 Sep 2026 06:46:09 GMT Docker Compose for ML projects: set up GPU-enabled multi-container development with health checks, watch mode, profiles, and resource limits. SageMaker Pipelines Tutorial: End-to-End MLOps on AWS https://samaustinai.pages.dev/2026/09/sagemaker-pipelines-tutorial-end-to-end https://samaustinai.pages.dev/2026/09/sagemaker-pipelines-tutorial-end-to-end Wed, 30 Sep 2026 06:33:58 GMT Build reproducible, auditable ML workflows on AWS: SageMaker Pipelines tutorial covering the @step decorator, step classes, and FailStep gates. Best RAM and Storage Upgrades for Local AI Workstations https://samaustinai.pages.dev/2026/09/best-ram-and-storage-upgrades-for-local https://samaustinai.pages.dev/2026/09/best-ram-and-storage-upgrades-for-local Tue, 29 Sep 2026 11:17:40 GMT Best RAM and storage upgrades for local AI workstations in 2026: how much memory 8B-14B models need, NVMe sizing, and how to shop through the DRAM shortage. Best Laptops for Running Local LLMs and Edge AI Development https://samaustinai.pages.dev/2026/09/best-laptops-for-running-local-llms-and https://samaustinai.pages.dev/2026/09/best-laptops-for-running-local-llms-and Tue, 29 Sep 2026 10:52:50 GMT Best laptops for local LLMs in 2026: why memory bandwidth beats GPU name, how much memory 7B to 120B models need, and Apple vs NVIDIA vs AMD picks. Speculative Decoding Explained: Speed Up Local LLM Inference https://samaustinai.pages.dev/2026/09/speculative-decoding-explained-speed-up https://samaustinai.pages.dev/2026/09/speculative-decoding-explained-speed-up Tue, 29 Sep 2026 10:34:35 GMT Speculative decoding explained: how draft models, n-gram lookup, Medusa, and EAGLE speed up local LLM inference 2-3x with mathematically identical output quality. Offline AI Apps: Building Machine Learning Apps That Work Without Internet https://samaustinai.pages.dev/2026/09/offline-ai-apps-building-machine https://samaustinai.pages.dev/2026/09/offline-ai-apps-building-machine Tue, 29 Sep 2026 10:30:37 GMT Build offline AI apps: how WebLLM, Transformers.js, and ONNX Runtime Web run machine learning in the browser with WebGPU acceleration and WASM fallback. Running Multiple Local LLMs: Model Switching and Routing https://samaustinai.pages.dev/2026/09/running-multiple-local-llms-model https://samaustinai.pages.dev/2026/09/running-multiple-local-llms-model Tue, 29 Sep 2026 07:33:42 GMT Run several local LLMs on one GPU: compare llama.cpp router mode, llama-swap, and LiteLLM for on-demand model switching, routing, and cloud fallback. Benchmarking Local LLMs: Tokens per Second Across Hardware https://samaustinai.pages.dev/2026/09/benchmarking-local-llms-tokens-per https://samaustinai.pages.dev/2026/09/benchmarking-local-llms-tokens-per Tue, 29 Sep 2026 07:30:19 GMT Measure real local LLM speed: tokens per second and time to first token across GPUs, Apple Silicon, and CPU, plus how to benchmark your own setup. Local Voice Assistants: Build a Privacy-First Alexa Alternative https://samaustinai.pages.dev/2026/09/local-voice-assistants-build-privacy https://samaustinai.pages.dev/2026/09/local-voice-assistants-build-privacy Tue, 29 Sep 2026 07:21:45 GMT Build a fully local voice assistant with Wyoming, Whisper, Piper, openWakeWord, Home Assistant, and Ollama — no cloud, no data leaving your home network. Apple Neural Engine Explained: Optimizing Models for Apple Silicon https://samaustinai.pages.dev/2026/09/apple-neural-engine-explained https://samaustinai.pages.dev/2026/09/apple-neural-engine-explained Tue, 29 Sep 2026 07:09:06 GMT How the Apple Neural Engine works, which operations actually reach it, how to check compute unit assignment in Xcode, and how to design Core ML models it runs. Edge Impulse Tutorial: No-Code TinyML for Embedded Devices https://samaustinai.pages.dev/2026/09/edge-impulse-tutorial-no-code-tinyml https://samaustinai.pages.dev/2026/09/edge-impulse-tutorial-no-code-tinyml Tue, 29 Sep 2026 06:31:51 GMT Train and deploy TinyML models without code using Edge Impulse: data collection, impulse design, the EON Tuner, validation, and deploying to Arduino or a browser. MediaPipe Tutorial: Real-Time On-Device ML for Vision and Audio https://samaustinai.pages.dev/2026/09/mediapipe-tutorial-real-time-on-device https://samaustinai.pages.dev/2026/09/mediapipe-tutorial-real-time-on-device Tue, 29 Sep 2026 06:05:16 GMT Run real-time vision and audio ML on-device with MediaPipe: setup for Python, Android, and web, gesture and audio pipelines, Model Maker, and pitfalls. Core ML Tutorial: Deploy Machine Learning Models on iPhone https://samaustinai.pages.dev/2026/09/core-ml-tutorial-deploy-machine https://samaustinai.pages.dev/2026/09/core-ml-tutorial-deploy-machine Tue, 29 Sep 2026 06:01:21 GMT Deploy machine learning models on iPhone with Core ML: convert PyTorch with coremltools, integrate in Xcode, run Vision predictions, and cut size. Fine-Tuning a Local LLM with LoRA on Consumer Hardware https://samaustinai.pages.dev/2026/09/fine-tuning-local-llm-with-lora-on https://samaustinai.pages.dev/2026/09/fine-tuning-local-llm-with-lora-on Mon, 28 Sep 2026 11:43:54 GMT Fine-tune a local LLM with LoRA and QLoRA on a consumer GPU: hardware sizing, Unsloth setup, dataset prep, training, evaluation, and running it locally. LM Studio Tutorial: Run LLMs Locally with a GUI https://samaustinai.pages.dev/2026/09/lm-studio-tutorial-run-llms-locally-gui https://samaustinai.pages.dev/2026/09/lm-studio-tutorial-run-llms-locally-gui Sat, 26 Sep 2026 10:14:00 GMT Install LM Studio and run LLMs locally: download quantized GGUF or MLX models, chat with documents offline, and serve an OpenAI-compatible API on your machine. Best Books on Deep Reinforcement Learning for Robotics https://samaustinai.pages.dev/2026/09/best-books-deep-rl-robotics https://samaustinai.pages.dev/2026/09/best-books-deep-rl-robotics Sat, 26 Sep 2026 10:06:00 GMT Four deep RL books compared for robotics work: Sutton & Barto, Lapan, Morales, and Kober & Peters — plus the reading order that actually works. Best Robotics Simulation Software Compared (MuJoCo, PyBullet, Isaac Gym) https://samaustinai.pages.dev/2026/09/best-robotics-simulation-software https://samaustinai.pages.dev/2026/09/best-robotics-simulation-software Sat, 26 Sep 2026 09:50:00 GMT Compare MuJoCo vs PyBullet vs Isaac Gym for robotics simulation: contact physics accuracy, learning curve, GPU-parallel RL training, and the Isaac Lab successor. Training AI Agents in Minecraft with Reinforcement Learning https://samaustinai.pages.dev/2026/09/training-ai-agents-minecraft https://samaustinai.pages.dev/2026/09/training-ai-agents-minecraft Sat, 26 Sep 2026 07:10:00 GMT Train AI agents in Minecraft with reinforcement learning: MineRL setup, VPT video pretraining, hierarchical RL, and Voyager LLM-driven planning. Path Planning with Reinforcement Learning for Mobile Robots: Mapless Navigation and Hybrid Architectures https://samaustinai.pages.dev/2026/09/path-planning-reinforcement-learning https://samaustinai.pages.dev/2026/09/path-planning-reinforcement-learning Sat, 26 Sep 2026 06:50:00 GMT Deep RL path planning for mobile robots: mapless navigation, costmap observations, reward shaping, dynamic obstacles, and hybrid RRT-plus-RL architecture. Training a Quadruped Robot to Walk with RL https://samaustinai.pages.dev/2026/09/quadruped-robot-walking-rl https://samaustinai.pages.dev/2026/09/quadruped-robot-walking-rl Fri, 25 Sep 2026 11:00:00 GMT Train a quadruped to walk with RL in Isaac Lab: 4,096 environments, terrain curriculum, domain randomization, teacher-student distillation, sim-to-real.