Michael Down

I architect agentic AI systems for regulated financial services: the layer that sits between AI agents and the systems they act on, and the controls that make their behaviour deterministic enough to run inside a bank. I have built production data platforms at bank scale, led the engineers who build them, and I still design and build myself.

I am currently working with a tier-one US bank to design the scope and architecture of an enterprise agentic knowledge layer, alongside the engineering team now taking it into production: a governed ontology spanning the bank’s processes, domain concepts, data products and physical systems, an MCP tool surface that agents call in order to act, platform-level trust enforcement that filters what any agent may retrieve by confidence and data sensitivity, and an execution-trace loop that turns reviewed runs into reusable paths.

I’ve spent over a decade in and around financial services data: real-time platforms at Barclays acting on billions of events a day, post-trade settlement under CSDR, then solution architecture at Elastic, then building Neo4j’s financial services business into the company’s largest industry vertical. I’ve worked with tier-one banks and insurers including Aviva and Zurich, and I own the vertical end to end: the technical direction alongside the commercial strategy.