I build intelligent systems inside the kinds of companies most AI commentary never visits — global enterprises with messy data, real customers, and SLAs that don't care about benchmarks.
For 25 years my work has lived on the integration layer: ERPs, CRMs, ITSM, financial systems. The unglamorous middle where models meet workflows, and where most enterprise AI projects either ship value or quietly disappear into a slide deck.
Today I lead enterprise applications engineering at N-able and teach machines to read service tickets, flag anomalies, predict churn, and triage incidents. Before that: Barclays Capital in Singapore, Ingram Micro, my own consulting firm, and a stretch in financial derivatives that I'm still finishing the PhD on — graph neural networks for option pricing.
This site is where I write down what I see. No tutorials. No "10 prompts that will change your life." Just notes from inside the room — what's working, what isn't, and why most of what you read about enterprise AI isn't quite right.
Designing and shipping LLM-powered assistants that triage real service tickets — not toy demos. Contextual response recommendation, knowledge base integration, escalation logic, and the unsexy plumbing that makes it all work in production.
In productionPhD research on graph neural networks applied to derivatives pricing prediction. Sitting at the intersection of my finance background, ML practice, and a long-running curiosity about whether structure-aware models can outperform classic numerical methods.
Active researchAnomaly detection across enterprise platforms — ticket volume spikes, SLA breach trends, resolution-time outliers. Building the data pipelines and models that move ops teams from reactive firefighting to predictive intelligence.
OngoingEmail is best. I read everything but reply selectively — usually within a week. Quickest way to get notified when I publish a note: drop me a line and ask.