Machine learning, shipped to production — since 2019

I'm Aaquib — an AI engineer who spent five years building production backends before specialising in machine learning. That order matters. I came to ML from systems work at Edcast, PropertyPistol and Norstella, where the job was making things fast, correct and still standing at 3 a.m. So when I train a model, my instinct is to ask how it gets deployed, what it costs per request, and how we'll know when it starts drifting.

Most AI projects don't fail at the demo. They fail three weeks later, in production, when nobody can prove the thing still works. My wedge is the unglamorous half — evaluation harnesses, regression suites, drift monitoring, citation grounding, and CPU inference paths that don't need a GPU bill to stay up.

I finished an MSc in Artificial Intelligence at the University of East London with a Distinction and a 93% average — top of my cohort, every module above 90 — and I now teach the Machine Learning and Big Data labs there. Outside work I ship small tools, contribute to the Tor Project and GitLab, and write at heyengineers.com.