May 2025 - March 2026
Principal AI Architect
Aveni - Edinburgh, UK
Architected the internal AI platform supporting enterprise banking deployments. Built Agent Assure and the evaluation, versioning and release controls around it.
The role covered the full stack of an AI-native transformation inside a regulated UK financial-services firm: agent runtime design, safety guardrails, and the governance case needed to bring agentic systems past a risk committee. Two things published elsewhere on this site came directly out of the work: the Tiered Governance Model, a 4-tier AI governance system aligned to NIST AI RMF, and a Trust Boundaries approach to managing risk across multiple LLM providers.
What I did
- · Architected the internal AI platform supporting enterprise banking deployments, aligning executive priorities, use cases, models, security, governance, operations and cost.
- · Built Agent Assure and designed workflows for call summarisation, conduct-risk scoring, AML/SAR signals, evidence generation, human review and escalation.
- · Established evaluation, prompt/model versioning, failure analysis, fallbacks, observability and release controls.
- · Applied the Tiered Governance Model and a Three Lines of Defence structure to agent deployments under SS1/23 and Consumer Duty.
FAQ
Where was Dipankar Sarkar the Principal AI Architect?
Dipankar was Principal AI Architect at Aveni in Edinburgh, from May 2025 to March 2026, working on regulated enterprise banking AI.
What regulatory framework did Dipankar apply at Aveni?
The Tiered Governance Model, a 4-tier AI governance system aligned to NIST AI RMF, paired with a Three Lines of Defence structure for SS1/23 and Consumer Duty compliance.
Did the work involve UK regulators directly?
Aveni's own announcement of 13 February 2026 confirms that Agent Assure completed a pilot in the FCA's inaugural Supercharged Sandbox. That was the company's programme, not an FCA approval; Dipankar can be specific in conversation about which parts of the work were his.
Related writing
- AI Architect in Financial Services: Lessons from Production
What I learned building production agent systems for financial services.
- AI Agent Governance for Financial Services
The Tiered Governance Model and Three Lines of Defence framework for SS1/23 and Consumer Duty.
- LLM Provider Risk Management for Banks
The Trust Boundaries Model and multi-provider fallback patterns for regulated industries.
- Tiered Governance Model
The published framework page: a 4-tier AI governance model for banks, insurers, and asset managers.