I have spent five years making one framework better

I am a Senior Software Engineer at QuantumBlack, AI by McKinsey, and a maintainer of Kedro, an open-source Python framework for building reproducible, modular data and machine-learning pipelines.

I joined as an intern in 2021, shortly before McKinsey donated Kedro to the Linux Foundation. I did the de-branding sweep across the organisation's repositories for that donation, which is a fairly unglamorous introduction to open source and a very good way to learn a codebase. Kedro has since graduated within LF AI & Data, and I sit on its Technical Steering Committee.

Most of my work is framework internals and developer experience: the parts users never look at directly, but feel immediately when they are wrong. I care about APIs that are hard to misuse, migrations that do not strand anyone, and error messages that point at the actual problem. Lately that has extended into the infrastructure around GenAI systems: tracing, evaluation and retrieval, on the argument that agents are the visible part of the problem and the pipeline is the hard part.

Before all of this I read Computer Science and then Artificial Intelligence at Queen Mary, where I co-authored a paper on real-time neuroevolution for game playing. Outside work I host the Kedro community's public technical stream.