Simon
Cullen

You spent five years at SIG. What came next?

Kraken — the digital asset exchange. A pivot into compliance. I built a compliance datamart and self-serve analytics platform serving regional compliance officers across multiple jurisdictions. Automated reporting processes that were taking days down to hours. Then OptumRx, where the scale is genuinely enterprise. I owned a 24/7 pricing ETL platform, led a full migration from R to Python and Databricks — significant performance improvements, real infrastructure cost reductions — and drove SOC 2 governance improvements alongside the technical work. In parallel since 2022 I've been running Seaduck Analytics, my own consultancy for data engineering and AI systems.

Tell me about the R to Python and Databricks migration at OptumRx. That sounds like a significant undertaking.

It was. The platform had grown organically in R over several years and it was starting to buckle under load. We rebuilt it in Python on Databricks. The performance improvements were significant. Infrastructure costs came down. The team could iterate faster. But the hardest part was never the technology — it was building confidence that the new system would be trustworthy on day one. You are replacing something mission-critical. Healthcare pricing. It cannot be wrong. I mentored the engineering teams through it and drove the SOC 2 governance improvements at the same time. That combination — technical leadership and process rigour, simultaneously — is what makes these migrations land safely.

You describe yourself as an AI-first engineer. What does that actually mean in practice?

It means AI tools are integrated into every phase of how I work — not just code generation, but ideation, testing, refactoring, documentation, and deployment decisions. I use Claude and agentic workflows daily. The speed change is real. Things that used to take a week can take a day if you are using these tools properly. I have shipped two production websites — seaduck.ie and this site, simon-cullen.com — designed and delivered using the latest AI agentic tools and workflows, from prototype through to deployment. That is not a parlour trick. That is the actual workflow now.

Beyond shipping sites, I run a separate experiment at insights.simon-cullen.com — a Ghost blog where an AI assistant manages the entire content pipeline. Research, drafting, formatting, image generation, audio production, publication — all handled by agents working largely unsupervised. It is a live testbed for understanding what breaks when you give AI real production responsibilities, and what actually works at scale.

Is the industry keeping up with that pace of change?

Some organisations are moving fast. Most are still working through the governance question — how do you use AI tools in a regulated environment? Healthcare, finance, compliance — you cannot just ship and iterate. You need rigour. That is actually the interesting engineering challenge right now: how do you bring AI-assisted development velocity into high-accountability environments without compromising safety? I am working through that every day at OptumRx and with Seaduck clients. The answer exists. It just requires more thought than most people are applying to it.

You hold several professional certifications. Which carries the most weight in today's AI-driven landscape?

Databricks Lakehouse Fundamentals resonates most right now. The lakehouse architecture — converging data lakes and warehouses — is where the enterprise industry has landed, and Databricks is at the centre of that for most large organisations. Azure DP-900 gives you the cloud data fundamentals that underpin almost everything in the Azure ecosystem. The FinOps certifications — Practitioner and now Engineer — are the ones that surprise people. Cloud cost management is not glamorous, but at enterprise scale the ability to reason about infrastructure spend separates a good data platform from an expensive one. More recently I completed the CCSK v.5 and CCZT from the Cloud Security Alliance, which reflects where the industry is heading — security and zero trust are no longer the infrastructure team's problem. They are the data engineer's problem too. But honestly? The most valuable credential in 2026 is not on any official list. It is the ability to ship an AI-integrated data product end-to-end. Certifications validate your baseline. Shipping validates your capability.

And the academic background — does an M.Sc in International Business translate into data engineering?

More than people expect. It gives you a frame for understanding why the data matters — not just how to move it. The business context behind a pricing system, a compliance requirement, a trading operation — that is what separates a data engineer who builds the right thing from one who builds a technically correct thing nobody needed. The Graduate Diploma in IT from DCU gave me the formal foundations. Together, they are genuinely complementary. Most data engineers can tell you how a system works. Fewer can tell you why it should exist and what it is worth. That gap is where I operate.

What is next for you, Simon? What are you building now, and what does the right opportunity look like?

What excites me right now is the bleeding edge — the problems most teams are still working out. Reinforcement learning applied to production systems. Model quantization and the real engineering trade-offs that come with deploying smaller, faster models without sacrificing capability. FinOps at AI scale, because the cost of running intelligent infrastructure is a genuine engineering problem, not just a finance conversation. And AI within cybersecurity — the attack surface is evolving faster than most defences, and that intersection is one of the most consequential engineering spaces right now. Ultimately I am drawn to the organisations defining what comes next — the labs and teams at the frontier of what models can actually do — and the data infrastructure challenges that exist at that level. More than any specific role, what I am looking for is creative people who are genuinely in the thick of these challenges. I want to be in those rooms. I want to hear what the hard problems actually are and work with the teams tackling them.