🤖 AI Stack

A living snapshot of the tools, reading, and learning that shape how I work with AI. Updated as my stack evolves.

Last updated: April 2026

Reusable prompt systems I've built for Claude. Each skill is a structured instruction set that changes how Claude approaches a specific type of task.

Pac-Man Explainer — Turns any complex technical or conceptual term into a short Pac-Man story narrative, then summarises the analogy in a comparison table. Use it when you need to explain something hard to a non-technical audience without dumbing it down. Trigger: “explain X using Pac-Man” or just drop a single complex term.

All skills are open and available on GitHub →

Claude — My primary reasoning partner for everything from data pipeline debugging to writing. If you do any technical work, this is the one to start with.
Gemini — Google's model, useful when I need a second opinion or want to cross-check Claude's output on factual queries.

Google Antigravity — Where I write all my code. It is built on VSCode. Google's agent-first IDE, still in preview. The Manager view for running multiple coding agents in parallel is genuinely new territory. Worth watching if you want to understand where agentic development is heading.
GitHub — Version control and portfolio. If you're building anything, you need this habit early.
Anaconda — Manages my Python environments and Jupyter notebooks cleanly. Saves you from dependency hell.

Meltano — Open-source EL pipeline tool. I use it to ingest data from public APIs into BigQuery without writing ingestion scripts from scratch.
dbt — Transforms raw data into clean, documented models inside the warehouse. SQL-first and version-controlled.
Apache Kafka — Handles real-time data streams. I use it in my air quality pipeline for architectural depth and skill-building.
Apache Spark — Distributed data processing. Useful when the data is too large for single-machine tools.
Docker — Containerises my full stack so it runs consistently across environments. Non-negotiable once you have multiple services running together.

Google Cloud / BigQuery — My primary cloud data warehouse. BigQuery's serverless model means I can run heavy queries without managing infrastructure.
AWS Lightsail — Hosts my web projects. I chose it over EC2 deliberately: I don't need EC2's full feature set, Lightsail does the job cleanly, and it costs less. Sometimes the simpler tool is the right tool.

AI Engineering by Chip Huyen — The most practical book on building with LLMs. Read this before anything else if you want to build, not just prompt.
Empire of AI by Karen Hao — A necessary read on the politics and human cost behind the AI industry. Skewed toward OpenAI but the dynamics apply everywhere.
Human Compatible by Stuart Russell — On my list. Russell argues we've been building AI wrong from the start. Feels important to sit with.
Deep Learning by Goodfellow, Bengio, Courville — The theoretical foundation. Dense, but free online and worth having as a reference.

AI & I by Dan Shipper — Thoughtful, not hype-driven. Good for thinking about AI and creative work without the noise.
AI Today — Practical and grounded. Good for staying current without doomscrolling.
Lex Fridman Podcast — Long-form conversations with scientists, engineers, and thinkers at the frontier. Not every episode is relevant but the ones that are, are worth every minute.

NTU Advanced Professional Certificate in Data Science and AI — Full programme covering Python, SQL, data pipelines, ML deployment, and generative AI. Currently completing this, finishing June 2026.
Anthropic Academy — Anthropic's own learning resources for building responsibly with Claude. Short, focused, and keeps you honest about what these models can and can't do.
NVIDIA Certifications — GPU computing and AI infrastructure. Useful for understanding what's happening under the hood when you deploy models at scale.

Research from labs and companies I track because they publish work that actually matters.

NVIDIA Research — Good explainers on AI infrastructure and what's coming at the compute layer.
Netflix Research — Surprisingly open about their ML systems. Real-world scale, real constraints.
Salesforce AI Research — I know this org well. They do serious applied research, not just product announcements.

You can read my AI posts for thoughts on how all of this plays out in practice.

Curious about the hardware and full environment behind this stack? See my Uses section.