Toolkit

I've spent 15 years building AI products and data teams. These are the tools, frameworks, and resources I actually use. Not a sponsored list, just what works.

GitHub

Open source projects, experiments, and the code behind some of the ideas I write about.

AI & Machine Learning

Python & PyTorch

Still the backbone. Most production ML work starts here. PyTorch for research and prototyping, ONNX for deployment where latency matters.

LangChain / LlamaIndex

For RAG pipelines and LLM orchestration. Good for prototyping agentic workflows, though I strip out the abstractions for anything going to production.

Hugging Face

Model hub, Transformers library, and Spaces for quick demos. The default starting point for most NLP and vision tasks before deciding whether to fine-tune or build custom.

Weights & Biases

Experiment tracking, model versioning, and artefact management. The thing that saves you when someone asks "which model run produced that result?" six months later.

Data & Infrastructure

dbt

Data transformation that actually scales. Brings proper software engineering practices (version control, testing, documentation) to the analytics layer where most teams need it most.

Snowflake / BigQuery

Cloud data warehouses that handle the heavy lifting. The choice between them usually comes down to your existing cloud provider, not the technology itself.

Airflow / Dagster

Pipeline orchestration. Airflow is battle-tested. Dagster is what I'd pick for a greenfield project. Either way, you need something that tells you at 3am what broke and why.

Docker & Kubernetes

Containerisation and orchestration for ML deployments. Not glamorous, but the difference between "it works on my laptop" and "it works in production."

Thinking & Strategy

Wardley Mapping

The best framework I've found for understanding where technology sits in its evolution and making build-vs-buy decisions that don't age badly.

RICE / ICE Scoring

For AI use case prioritisation. Simple frameworks that force honest conversations about impact, confidence, and effort before you've spent six months building something nobody wanted.

Team Topologies

The book that changed how I think about org design. Stream-aligned teams, platform teams, and the interaction modes between them. Required reading for anyone scaling a tech org.

The Lean Startup / Shape Up

Lean for the mindset, Shape Up (Basecamp) for the execution mechanics. Together they form the basis of how I run innovation programs inside large organisations.

Reading List

Books and papers that shaped how I think about AI, teams, and building things that last.

Designing Machine Learning Systems

Chip Huyen

The best book on ML systems in production. Covers the 90% of ML work that isn't model training: data pipelines, feature stores, monitoring, and all the infrastructure that actually makes models useful.

An Elegant Puzzle

Will Larson

Systems thinking applied to engineering management. The chapter on organisational design alone is worth the price. I've given this book to every new leader on my teams.

Thinking in Bets

Annie Duke

Decision-making under uncertainty. Changed how I approach AI strategy. Most decisions are bets, and the quality of the decision isn't determined by the outcome.

Attention Is All You Need

Vaswani et al., 2017

The transformer paper. If you work in AI and haven't read the original, you're building on a foundation you don't understand. Still the most important paper of the last decade.

Want to go deeper?

I write about how these tools and frameworks play out in practice in my weekly newsletter.

Read Maligned