Michael Down
Michael Down
About
Detail
Global Head of Financial Services
United Kingdom
I architect the agentic AI systems the world's largest banks trust with financial crime, risk and regulatory compliance, and I still build them myself.
As Global Head of Financial Services at Neo4j I own the company's largest industry vertical end to end: the technical direction alongside the commercial strategy, the category narrative now used across tier-one engagements, the public use-case library the field runs on, and the product and acquisition strategy behind it.
The work is concrete. I'm currently working on the knowledge layer a tier-one US bank is taking into production: a governed ontology over the bank's processes, data products and physical systems, an MCP tool surface agents call in order to act, and platform-level trust enforcement that decides what any agent may retrieve. Benchmarked with a top-three UK bank's audit team, it returned 98% retrieval accuracy across 250 questions with zero false positives, at 60% lower token cost.
I've spent over a decade close to these problems: real-time security platforms at Barclays acting on billions of events a day, post-trade settlement under CSDR, and financial crime and transaction monitoring architectures that run in production at tier-one banks today. I know not just how to build the solution, but how a regulated bank actually gets permission to run it.
I also put a point of view into the market: quoted in the Financial Times on AI and financial crime, and founder and host of GraphTalk, an invite-only forum for 30 to 40 senior financial services leaders in Manhattan.
Deep domain expertise in financial crime and risk. The regulatory fluency to navigate real supervisory environments. The architecture to ship at bank scale. That combination is what moves an institution from GenAI experimentation to governed production deployment.
As Global Head of Financial Services at Neo4j I own the company's largest industry vertical end to end: the technical direction alongside the commercial strategy, the category narrative now used across tier-one engagements, the public use-case library the field runs on, and the product and acquisition strategy behind it.
The work is concrete. I'm currently working on the knowledge layer a tier-one US bank is taking into production: a governed ontology over the bank's processes, data products and physical systems, an MCP tool surface agents call in order to act, and platform-level trust enforcement that decides what any agent may retrieve. Benchmarked with a top-three UK bank's audit team, it returned 98% retrieval accuracy across 250 questions with zero false positives, at 60% lower token cost.
I've spent over a decade close to these problems: real-time security platforms at Barclays acting on billions of events a day, post-trade settlement under CSDR, and financial crime and transaction monitoring architectures that run in production at tier-one banks today. I know not just how to build the solution, but how a regulated bank actually gets permission to run it.
I also put a point of view into the market: quoted in the Financial Times on AI and financial crime, and founder and host of GraphTalk, an invite-only forum for 30 to 40 senior financial services leaders in Manhattan.
Deep domain expertise in financial crime and risk. The regulatory fluency to navigate real supervisory environments. The architecture to ship at bank scale. That combination is what moves an institution from GenAI experimentation to governed production deployment.
