About Me

Leon

In my daily work, I build enterprise end-to-end machine learning solutions, including full data pipelines, training pipelines, and deployment pipelines.

This site is a structured knowledge base for practitioners: concept-first explanations, implementation details, and engineering tradeoffs drawn from hands-on work and course-driven study.

Core Expertise

• Machine Learning and Deep Learning: model fundamentals, optimization, and production-oriented reasoning.
• LLMs and Generative AI: alignment, post-training, and modern model architecture analysis.
• Reinforcement Learning: beginner foundations through policy optimization and LLM-RL connections.
• Deep Learning Systems and GPU Acceleration: autograd internals, systems design, and Triton-level kernels.
• Technical Writing: long-form notes designed to help readers move from theory to implementation.

Recommended Reading

Reinforcement Learning for Beginners

• Reinforcement Learning Basics
• Reinforcement Learning Introduction
• Imitation Learning

Build Deep Learning/PyTorch-Style Foundations

• Implement Your Own Deep Learning Library using Automatic Differentiation I
• Fully Connected Networks, Optimization, Initialization and Activations

GPU Programming and Acceleration

• DLSys Hardware Acceleration
• Triton Introduction

DeepSeek Series

• DeepSeek Base Models Series
• DeepSeek Reasoning Models Series

LLM + RL

• Reinforcement Learning (RL) From Fundamentals to PPO and GRPO in LLMs (I)
• Reinforcement Learning (RL) From Fundamentals to PPO and GRPO in LLMs (II)

Beyond Work

Outside AI, I enjoy sports, music, photography, and travel. Photos used on this blog are from my Unsplash and Pexels pages.