Headline

Senior AI PM| Agentic Workflows · LLM Evals · Enterprise Platform · GenAI · RAG | 30% Call Deflection @ Zeta

About

I build GenAI products that hold up in production. Not just demos. My craft is agentic AI systems and LLM-powered platform experiences, grounded in structured customer discovery: deciding when AI is the right call, choosing the right architecture (RAG vs. agentic flows vs. hybrid NLU+LLM), designing human-in-the-loop (HITL) guardrails and evaluation frameworks to know it's actually working, and driving 0-to-1 launches that scale under real enterprise load. At HashedIn by Deloitte, I define the product roadmap for AI-powered enterprise solutions: scoping agentic use cases, applying build-vs-buy frameworks, and partnering with engineering to prototype before a line of production code is written. Before that, at Zeta, I led the 0-to-1 build of an Agentic AI Conversational Platform for banking and fintech - built on a three-tier orchestration model: RAG self-serve, agentic transactional flows, and human escalation. That meant analyzing 100K+ call transcripts via LLM clustering to surface 69 distinct contact reasons, running GPT-4o vs. Claude evaluations, fine-tuning to 93.5%+ command generation accuracy, and building a custom evaluation framework adopted across the platform. Outcome: ~30% call deflection, 30% faster MVP launch. I partner cross-functionally across engineering, applied science, and GTM - using SQL-based analysis and structured customer signal to keep roadmap decisions grounded in evidence, not instinct. What I bring to AI platform roles: End-to-end system thinking: architecture tradeoffs, eval design, production guardrails, and HITL escalation Evaluation depth: component-level evals across ASR, NLU, and LLM command generation - I treat evals as a first-class product Hands-on prototyping: working demos before engineering scopes a sprint. Vercel deployments, agentic co-pilots, competitive research tools Enterprise platform judgment: build-vs-buy rigor, cross-system integration, NFR definition at scale Earlier: founding PM for a $26M+ B2B SaaS device platform at BYJU's (Osmo, Tynker, Epic - 100M+ learner audience), enterprise data intelligence at Accenture ($4M+ in client savings), software engineering at S&P Global. CS from VIT | MBA from SP Jain | Exchange at CMU Tepper. I'm drawn to roles where AI product decisions involve real technical tradeoffs - where the question isn't "should we use AI?" but "which architecture, what eval threshold, and what does success look like?"

Location

Bengaluru

AI stack

claude code

ChatGPT

Work experience

Vice President - Product Manager at JPMorganChase, Jul 2026 - Present · 2 mos, Bengaluru, As a Product Builder, manage the suite of agentic solutions to tackle problems related to improving the effectiveness and controls of the Product Development Life Cycle.

Senior Product Manager at HashedIn by Deloitte, Feb 2026 - Jul 2026 · 6 mos, Bengaluru, Defining the product roadmap for AI Coworkers: Claude-powered agentic solutions that automate SDLC workflows end-to-end for client organizations. Partnering with engineering and delivery leads to prototype proof of concepts reducing engineering scope ambiguity and de-risking build decisions. applying build-vs-buy frameworks to scope, prototype, and scale.

Independent AI Product Builder at Self-Employed, Aug 2025 - Jan 2026 · 6 mos, Remote, Designed and co-built Nexus Fleet Intelligence, an agentic AI co-pilot for trucking load-bidding that reduces bid decision time from ~10 minutes to seconds. Recognized as most market-ready idea at Maven AI PM Community Pitch Day 2025. Built and deployed a company and competitive landscape research AI agent, live at v0-competitor-analysis-tool-pi.vercel.app. Hands-on 0-to-1 product build from problem definition to production deployment.

Senior Product Manager at Zeta, Mar 2024 - Apr 2025 · 1 yr 2 mos, Bengaluru, Led 0-to-1 build of an Agentic AI Conversational Platform (voice + chat) for banking/fintech enterprises, deployed on AWS + Azure serving 150K+ customers with ~30% call deflection. Analyzed 100K+ call transcripts via LLM clustering to surface 69 distinct contact reasons, quantify volume share, and rank automation ROI, driving a focused MVP roadmap and reducing engineering scope ambiguity. Fine-tuned GPT-4o to 93.5%+ command generation accuracy after head-to-head benchmarking vs. Claude; built a custom evaluation framework adopted across the AI platform for ongoing solution validation. Designed and executed component-level evals across the full pipeline: ASR (WER, latency <400ms), NLU classifiers (95% on GPT-generated data, 83% on real transcripts, improved 2% via human-in-the-loop). Architected hybrid NLU + LLM system design with guardrails for hallucination, escalation triggers, and multi-intent fallback handling. Three-tier resolution: RAG self-serve, agentic transactional flows, human escalation. Enabled 30% faster MVP launch through assumption-first prioritization, stakeholder alignment across business and engineering, and lean validation cycles.

Senior Product Manager at Osmo, Sep 2019 - Feb 2023 · 3 y 5 mo, Bengaluru

Product Manager at Accenture, May 2018 - Aug 2019 · 1 y 3 mo, Bengaluru

Software Engineer at Capital IQ, Jul 2013 - May 2016 · 2 y 10 mo, Gurgaon

Education

Master of Business Administration (M.B.A.), Information Management, SPJIMR SP Jain Institute of Management & Research

Workplace Communication in Canada for IT Professionals, Business Communication, Toronto Metropolitan University