How I Build AI Agents From Masterclass Guides, Not Prompt Templates
A practical pipeline for turning expert judgment into reusable agents and skills that survive real work.
Writing Archive
Blog
Notes on production AI systems, agent design, prompt optimization, and the machine learning plumbing that decides whether ideas survive contact with real users.
Archive
6 pieces, newest first.
Programmatic Prompt Optimization: Building a Spam Filter with DSPy and MIPROv2
Why prompt engineering by intuition stops working in production, and how DSPy plus MIPROv2 let you optimize prompts against explicit metrics.
Do You Know What Agents Are?
A practical definition of AI agents, what actually counts as agentic, and when simpler LLM systems are the better engineering choice.
Running State: Enhancing Short-Term Memory in AI Agents
How agents carry context through complex tasks, and why memory design matters as much as the model choice.
scikit-learn Pipelines
How to bundle preprocessing and model training into reusable scikit-learn pipelines without letting feature engineering sprawl.
Tabular Q-Learning
A hands-on introduction to tabular Q-learning, the Bellman equation, and training an agent on Frozen Lake.