Work

Projects & case studies.

A detailed look at the products I've led — what the problem was, how I approached it, and what happened.

Product Manager, AI/Platform · HCLTech · December 2022 – December 2024

HCL Intelligent Operations

Enterprise AIOps platform — 140+ clients, three-time analyst Leader

End-to-end AIOps platform spanning alert correlation, root cause analysis, self-healing automation, and intelligent Service Management. I owned the product roadmap from discovery through GTM, working closely with data science, platform engineering, and a global sales team to drive Net New and Existing New bookings for 140+ enterprise accounts across varied domains.

→$17M+ incremental revenue over three years

→140+ enterprise clients

→Gartner, IDC & ISG Leader — three consecutive cycles

→8% MTTR reduction for top 10 accounts

→25% reduction in customer churn

Case study

The problem

Enterprise IT operations teams were drowning in alert noise. Thousands of events per hour, manual triage, and no systematic way to connect alerts to underlying causes — MTTR was high and ops cost was climbing.

What I did

I led roadmap decisions across four product areas: alert correlation (ML-based noise reduction), root cause hinting (LLM-assisted diagnosis), self-healing automation (runbook execution), and Agentic Service Management (end-to-end L1/L2 ticket handling without human intervention). Each area was prioritised using RICE and validated against a core group of 10 design-partner accounts.

The outcome

$17M+ incremental revenue over three years. Three consecutive Gartner, IDC, and ISG Leader ratings. 25% reduction in client churn. For our top 10 accounts: 8% MTTR reduction and 14% lower ops cost.

Stack & methods

Azure OpenAILLM OrchestrationITSM IntegrationMulti-tenant SaaSPython

Product Manager · HCLTech · January 2025 – December 2025

AIForce.Ops

Agentic AI framework for autonomous IT operations

Led AIForce.Ops, an Agentic AI framework purpose-built for IT operations. The product enables auto-triage, runbook execution, and self-healing for L1/L2 incidents without human intervention. I was a key part of this initiative, monitored directly by the CEO's office, working closely and managing the expectations of a 25-person cross-functional team, aiming to be the first organisation to get a unified agentic solution from zero to market-ready, including a brand new interactive demo environment that superseded the first one.

→40% reduction in manual interventions (early-adopter accounts)

→$5M in net-new bookings from demo environment

→30+ tenant phased rollout

→25-person cross-functional team

Case study

The problem

Even with AIOps reducing noise, the last mile — actually resolving incidents — still required human intervention. L1/L2 analysts spent the majority of their time on repetitive, well-understood ticket types that followed predictable resolution paths.

What I built

An agentic framework where LLM-powered agents receive incident context, select the appropriate runbook, execute remediation steps via tool calls (API, shell, ITSM), and close tickets with an audit trail. I defined the agent architecture, evaluation metrics (latency, hallucination rate, task success rate), guardrails, and the phased rollout strategy across 30+ tenants.

The outcome

40% fewer manual interventions on early-adopter accounts. The interactive demo environment — built with synthetic data and scripted agent workflows — was used in 100+ sales pitches and correlated with $5M in net-new bookings over 12 months.

Stack & methods

Agentic AIAmazon BedrockAzure OpenAITool OrchestrationPython

PM Track · Mobikwik · December 2021 – June 2022

Mobikwik BNPL Onboarding

Redesigned credit eligibility UX and onboarding flow on consumer fintech app, leveraging Clevertap funnel data. Cut sprint spillovers by 65% by restructuring backlog grooming and aligning PM–engineering on acceptance criteria.

→+12% BNPL sign-ups (50K incremental QoQ)

→Funnel + cohort dashboards owned end-to-end

Case study

The problem

BNPL onboarding had a steep drop-off at credit-check where users who didn't understand why they were rejected or what they could do about it, would just stay away from the platform.

What I did

Funnel-first analysis in Clevertap identified the exact drop-off point. Qualitative interviews with rejected users surfaced the confusion. I redesigned the eligibility step with proactive disclosure (what we check and why) and a clear path back into the funnel. Shipped behind a 50/50 split with primary metric on activated accounts.

The outcome

12% relative lift in completed BNPL sign-ups — 50K incremental registrations QoQ. Decision time on funnel issues fell 50% once we standardised cohort dashboards across the team.

Stack & methods

ClevertapFunnel AnalyticsA/B TestingMobile Onboarding

Want to talk through one of these projects?