Senior AI Product Manager, Regulated Agentic Systems, Ascendion
Mar, 2026 - Present
Embedded as sole PM on an internal agentic product being piloted with a regulated banking client. The product auto-generates epics, features, and user stories end-to-end from a high-level initiative brief, targeting production-ready output including code and tests.
Translated 15 UK banking regulations including FCA Principles, BCOBS, PSR 2017, and the Data Protection Act; into structured story templates that encode compliance requirements as a first-class input to AI-generated artifacts, replacing human-dependent recall with systematic coverage across all future projects.
Designed and implemented a 9-point rubric-based evaluation system for scoring agentic output quality across PRDs, features, stories, and tasks enabling comparative measurement across prompt sets and establishing a repeatable standard for production readiness.
Reduced token costs and improved agent focus through context partitioning: splitting instructions across role-specific files so each agent receives only relevant context, replacing prompt verbosity with targeted retrieval.
Designed tiered model routing across the agentic pipeline, matching task complexity to model capability to reduce inference cost without sacrificing output quality.
Designed and implemented pre/post execution hooks within the agentic pipeline including linting, validation gates, and output checks; both independently and in collaboration with engineers on a custom framework.
Each initiative generates 40–50 stories; the system reduces PM delivery time by approximately 10x, enabling a single person to cover both PM and TPM scope.