About

I'm Sagar Bharambe, a senior risk and governance leader based in the Netherlands. I build AI-powered multi-agent systems for financial resilience, regulatory intelligence, and human-centric governance.

I spent years in governance and risk roles across financial services, watching institutions struggle with the same problems: fragmented regulatory data, siloed risk assessments, and the persistent gap between compliance frameworks and operational reality. That experience convinced me that the problem is architectural, not analytical. Most organisations have the expertise. What they lack is the structure to make that expertise visible where it matters.

My work draws heavily on Phanish Puranam's research on organisational design at INSEAD. Puranam frames every coordination problem, human or algorithmic, through four universal challenges: how work is divided, how it is allocated, how quality is incentivised, and how information flows between actors. That framework shapes how I design multi-agent systems. Agents are not just software components. They are members of an organisation, and the same design principles that determine whether a human team produces genuine insight or compliance theater apply to teams of AI agents.

That conviction runs through everything I build. Roach Resilience deploys specialised AI agents that assess cascading risks across geopolitical, financial, regulatory, and operational domains for Dutch financial institutions, preserving disagreement between analytical frameworks rather than averaging it away. Eryndal indexes over eight million regulatory and security nodes with structure-aware chunking, making regulatory landscapes navigable through engineering. MAGS is my research into multi-agent governance itself: the organisational theory, the architectural patterns, and the open problems that determine whether these systems produce real oversight or just the appearance of it.

I believe the best risk systems make complexity legible without pretending it is simple. The goal is never to reduce uncertainty to a single number. It is to give decision-makers the structure to reason about it well.