Product Manager with 6+ years across fintech payments, enterprise data transformation, and AI systems — with a track record of owning delivery from spec to production. Built a live Generative AI (GenAI) evaluation and observability pipeline from scratch: designed the tracing layer, defined LLM eval criteria, and cut system costs 85% on a live product. Earlier, managed a white-label remittance platform serving regulated UK and Singapore banking markets as Associate PM — owning change requests, UAT triage, incident RCA, and production releases. At EY, delivered programme governance across three concurrent Unilever global transformation workstreams at Fortune 500 scale. Strong bias toward shipping, systems thinking, and AI-first product development.
EY (Ernst & Young)
Delivery Excellence · Client: Hindustan Unilever Limited (HUL), Fortune 500 Managed programme governance across 3 concurrent Unilever global transformation workstreams — U2K2 (SAP ERP standardisation), Project Sirius (supply chain harmonisation), and SAP BTP (cloud extension migration) — consolidated inputs from 6+ workstream leads into executive steering dashboards consumed by C-suite programme sponsors Served as single point of coordination across 3 organisations — TCS (migration), Infosys (development), and HUL IT — for cloud migration of safety and health applications from on-premise and GCP to Azure; tracked cross-vendor dependency chains and surfaced blockers before escalation thresholds Ran demand intake and feasibility triage for PACE enterprise automation programme — managed inbound Power Automate requests from cross-functional HUL teams, assessed build-readiness, and tracked concurrent delivery across multiple automation owners Diagnosed root cause of stalled ML asset-classification initiative (RTV): traced model failure to upstream SOP non-compliance corrupting training data — documented findings, escalated to programme leads, recommended project pause; recommendation accepted and implemented
Avenues Payments India (CCAvenue)
White-label Remittance Software · UK & Singapore Markets · 10K+ Transactions/Day Owned end-to-end change request lifecycle for a regulated payments platform — translated client-submitted requests into prioritised, signed-off specification documents across sprint cycles; zero unscoped changes entered active builds without formal CR documentation Managed production deployments on a live payments platform — authored release notes, confirmed cross-functional availability across client, tech, and database teams; maintained zero-tolerance rollback protocol on all failed patches, coordinating deployments routinely until 2AM Triaged all UAT findings as defect, scope creep, or new requirement — proved contested defects live in front of clients; prevented unscoped work from inflating sprint scope across all delivery cycles Prioritised and routed compliance changes across 2 regulatory frameworks — CHAPS (UK) and MAS (Monetary Authority of Singapore) — ensuring correct platform received each regulatory fix with no cross-market contamination Designed end-to-end user journey wireframes for mobile application; presented to product leadership and secured sign-off — wireframes adopted as product blueprint; mobile app development initiated post-handoff based on this foundation
Bigbasket.com
Reduced inventory write-offs ~35% by building detection logic for location-level losses and bin-level mismatches flagged during warehouse audits — identified process gaps and drove compliance corrections across affected locations Improved product availability and cut undelivered orders by analysing fill rate failures — traced root causes across stockout patterns, bin allocation errors, and planogram non-compliance; recommended and tracked corrective actions across warehouse teams
MBA
Shipped a live LLM evaluation pipeline end-to-end — defined eval criteria across cost, latency, and accuracy; traced routing decisions across multi-step model calls; iterated on routing logic based on observed failure patterns Designed and defended API contracts across 3 engineering iterations: input schema (intent class, token budget, fallback priority), output schema (selected model, confidence, routing path) Cut costs ~85% and latency ~81% vs. unoptimised baseline — ~$53K annual savings at 10K queries/day through eval-driven product iteration on a live system Built telemetry pipeline from scratch: captured decision traces, aggregated cost and latency per intent class, surfaced production anomalies
Designed architecture serving 3 stakeholder tiers from a single governed dataset — eliminated 3 redundant reporting workflows. Defined data contracts between source layer and each consuming dashboard surface; reduced reporting lag and prevented integration rework
Built probabilistic behaviour modelling and automated gap detection against historical survey data Building eval layer to validate AI-generated survey quality against human-reviewed baseline — same scaling challenge as enterprise LLM evaluation systems
Airtribe
Wharton Online