Open source

Tools, experiments and lessons in code.

I build small tools to explore AI engineering problems: choosing models, understanding costs, improving retrieval and connecting tools.

My GitHub profile

Selected projects

Featured project

model-switcher

Python · MIT licence

Model routing and cost tracking

Routes Claude Code prompts between model tiers and tracks turn and session costs. An offline heuristic scores each prompt, sending simpler work to a lower-cost model and harder work to a stronger subagent. Routing decisions and cost calculations run locally.

View model-switcher on GitHub

claude-session-analyser

JavaScript

Understand coding-agent costs

Analyse Claude Code session costs, forecast spending and find opportunities to save tokens. Session analytics run on your machine.

View session analyser on GitHub

myautorag

Python

Build and evaluate a retrieval pipeline

A command-line toolkit for building and evaluating retrieval-augmented generation pipelines from PDFs. It explores chunking, embedding, retrieval, reranking, generation and evaluation.

View myautorag on GitHub

simple-math-weather-servers

Python

Learn how MCP tools connect

A Python example connecting an agent to maths tools through the Model Context Protocol (MCP), with an optional weather server and troubleshooting notes.

View MCP examples on GitHub

More exercises, coursework and experiments.

Browse all repositories