Experience & learning.

Agent-Assist Next Best Offer Engine (Insurance): Architected a recommendation engine combining a Neo4j persona knowledge graph with lifetime-value, likelihood-to-act and product-fit models — served from AKS via Helm with Azure Blob-backed model artifacts.

Legacy Modernization — COBOL/Tandem to .NET on Azure: Leading discovery and phased migration strategy for a mainframe-to-cloud program — AST-driven transformation pipeline, spec-driven development with AI coding agents, and golden-master validation to guarantee functional equivalence.

Executive Advisory & Technical Presales (BFSI): Primary AI technical architect for global BFSI business lines — led design workshops and presented GenAI ROI and architecture strategy directly to executive leadership.

Prioritization Rubrics across Business Units: Built and executed a standardized evaluation rubric to prioritize architecture requirements across multiple business units in parallel, matching recommendations to each unit's engineering maturity and business velocity.

Enterprise-Scale RAG — DB Omni: Implemented an organisation-wide RAG pipeline integrating GitHub, Bitbucket and Confluence into a single unified search; enabled code explainability with business context and surfaced knowledge gaps — materially reduced information-search time.

Agentic QA & Reconciliation Automation: Architected an enterprise multi-agent system (LangGraph, LangChain) automating end-to-end test execution and post-trade data reconciliation across banking platforms — QA testing cycles reduced by ~60%.

Corporate Cashflow Analytics — Deterministic ML at Scale: Implemented agentic workflow using deterministic tool calls and a custom-trained model with confidence scoring; low-confidence cases route to human review, feeding continuous retraining. Processes 10B+ monthly records — ~70% reduction in compliance rule-change effort, ~30% performance improvement.

Regulatory Data-Adjustment Automation: Automated manual data adjustments in the back-office regulatory reporting warehouse (event-driven architecture on GCP — BigQuery, Dataflow — and AWS), reducing Functional Analyst adjustment effort by ~70%.

Enterprise LLMOps: Established distributed tracing and audit logging so every automated financial decision is traceable; designed functional evaluation metrics for real-time model and agent quality.

Back-Office Position Management (Capital Markets): Built core backend for an enterprise transactional Data Warehouse managing real-time account-position changes across global securities infrastructures; used ESB to handle high-frequency capital-markets data.

Dodd-Frank Compliance & Derivatives Processing: Engineered middle-office integration platforms handling FIXML APIs for DTCC and MarkitWire to process Credit Default Swaps (CDS) and Interest Rate Swaps (IRS) trade data for Dodd-Frank regulatory reporting.

Trading Platform Integrations: Built secure, low-latency "Smart Client Frameworks" to route real-time transaction data from Bloomberg front-office terminals downstream into Global Calypso back-office systems for post-trade clearing and settlement.

Credit Bureau Bulk Ingestion (Consumer Credit / Fintech): Designed high-volume, secure ETL pipelines ingesting bulk financial data from Micro-Finance Institutions; automated cleansing and transformation into regulator-mandated formats for credit-bureau reporting.