Practices

Pages

  • Code World Model
    Overview of Meta's Code World Model (CWM), a 32B-parameter language model that simulates code execution, with benchmark results and download links.
  • DoorDash Feature Store Redis Optimization
    How DoorDash benchmarked key-value stores and tuned Redis with hashes, xxHash, protobuf, and Snappy to cut its ML feature store's cost, memory, CPU, and latency.
  • Human Action Recognition
    Explains why combining 3D-CNN with LSTM handles long, multi-stage human actions in video better than either model alone.
  • Pose Estimation
    Introduces pose estimation, the task of locating human body keypoints in images and video, and the popular YOLO-Pose and MediaPipe Pose solutions.
  • Queuing Theory In Model Serving
    A debugging story showing how modeling replayed traffic as a Poisson process, rather than fixed intervals, explained a 2x p99 latency gap in GPU model serving.
  • RAG Tutorial Neo4j GraphRAG
    Step-by-step tutorial on building a RAG system on a Neo4j knowledge graph with LangChain, plus best practices and common pitfalls for GraphRAG.
  • Recommender Systems
    Overview of recommender system design, from a two-tower four-stage architecture to an AWS Personalize and generative-AI marketing solution, with a comparison of the two approaches.
  • Senator Knowledge Gap Detection
    Explains SENATOR, a framework that uses structural entropy, a knowledge graph, and MCTS to find LLM knowledge gaps and patch them with targeted synthetic-data fine-tuning.
  • TDD With Coding Agents
    Why forcing strict test-first TDD on a coding agent often costs more than it returns, and what to do instead - outcome feedback loops, regression tests, static-analysis feedback tools, and ATDD on a hexagonal or vertical-slice architecture.
  • TikTok Like Recommender And AWS Personalize
    Walkthrough of a TikTok-style two-tower recommender for H&M and an AWS Personalize plus Bedrock ad-copy solution, with a comparison of when to use each.
  • TikTok Strategy
    A first-person account of three AI-assisted workflows (research, retention optimization, and a 30-day content factory) for growing a TikTok channel quickly.