Suranga NanayakkaraI kept following the problem
one layer down.

I build data, software and AI systems. Not one of those three, all of them, because in my experience the interesting failures always happen where they meet.

I started in 2015 writing mobile and desktop applications for an insurer. Ordinary software engineering. The thing that pulled me away from it was noticing that the data those applications produced was nobody’s responsibility. Reports were rebuilt by hand every month, numbers disagreed depending on who you asked, and none of that was going to be fixed by writing a nicer screen.

Down a layer, then another

So I moved into data engineering: pipelines, warehousing, the plumbing. Then it became clear the pipelines existed to feed models, so I moved into machine learning, and built predictive systems that had to score a customer while a sales adviser was still on the phone with them. Then into data science proper, leading teams doing forecasting and anomaly detection at telco scale. And now AI development, which in practice means retrieval and agent systems sitting on governed data.

Every one of those moves was the same move. Whatever I was building, the problem worth solving turned out to be one layer underneath it. That is the whole career, honestly. It also means I am hard to categorise, which has cost me a couple of job applications and won me most of the interesting work.

What I am actually like to work with

Most of my client work is unglamorous: taking something important out of a system nobody wants to touch, and doing it without breaking the business sitting on top. The technology is rarely the hard part. The hard part is proving the new numbers match the old ones well enough that somebody is willing to sign off on switching the old thing off.

I do not name clients on this site. You can usually guess, and that is fine, but writing the logos out felt like taking credit for other people’s brands.

Teaching

I lecture at university level. I have given guest lectures on artificial intelligence at Malmö University and taught artificial intelligence, business intelligence and blockchain to students at Cardiff Metropolitan University. I also speak at Azure and Databricks community events, and I am accredited as a Microsoft Certified Trainer.

I am not doing this to build a personal brand. I do it because explaining something to a room of people who are allowed to interrupt you is the fastest way to discover which parts you only thought you understood. The 12 certifications are the same instinct in a more boring format.

After hours

I build small systems that run without me. A scraper that watches job boards overnight and scores each posting against a CV with a language model. A watcher that polls housing platforms and alerts once per listing. Nobody asked for either. They are how I learn, and most of what I know about keeping something alive in production came from things that would only ever embarrass me if they broke.

If you want to work together, the short version is here. Otherwise I am onLinkedIn andGitHub, or justemail me.