





















I'm a Technical Product Manager with 5+ years bridging business and technology teams to ship AI-first, data-driven platforms. My work spans LLM integration, RAG pipelines, agentic workflows, and API-driven integrations — always with a prototype-and-iterate mindset. I'm currently filing a patent with my CTO for a 5-layer multi-agent AI platform, and I hold a Masters in Engineering Management from Cornell (3.88 GPA, full merit scholarship) with a DBA in progress at Westcliff University.
These automations run in production inside the Global Trade Platform — cron jobs, Jira and Freshservice syncs, and a RAG-powered BRD scanner. Each shows its actual node pipeline.
Monitors website health across GTP client environments. Add URLs, crawl every linked page, detect down pages and broken links. Runs on a scheduled cron and surfaces reports with flagged links for the delivery team.
Open live app ↗Pipeline pulse for DPS enhancements built for leadership. Pulls tickets from Jira and Freshservice from Jan 2025 onward, computes backlog health, monthly flow, and release-train mapping, with searchable ticket catalogs.
Open live app ↗Self-serve device diagnostics web app that runs an in-browser health check across battery, network, sensors, and storage. Generates a shareable report agents attach to claims — cutting back-and-forth on no-fault-found returns.
Open live app ↗A single portal to look up product intake requests — status, timing, linked Jira updates, and stakeholder comments in one place. Replaces scattered email threads and manual Jira checks for request submitters.
Open live app ↗A lightweight web app for CX agents to look up customer device coverage, claim eligibility, and entitlement in seconds. Replaces three legacy tools with a single search box backed by the GTP entitlement API.
Open live app ↗Most PMs ask 'what should the AI do?' I ask 'what happens when it's wrong?' In regulated domains like insurance and healthcare, wrong answers are compliance violations, not bad UX. Designing guardrails IS the product work.
In regulated environments, adoption flips the moment a human stays in control. My Traceability agent only took off once I added the sign-off loop. Agents do the heavy lifting; people make the call with a visible audit trail.
A working demo beats any deck. I use Lovable, v0, and Claude tooling to turn abstract LLM orchestration into interactive products — which is how I've secured executive sponsorship for AI-first initiatives.
Battle-tested PM patterns from regulated-domain AI work. Each one earned its spot by surviving production.
Nikhar takes the messiest, most ambiguous AI problems and turns them into product roadmaps engineering can actually build. She is the rare PM who can debate a vector-store choice in the morning and a quarterly OKR in the afternoon.
She shipped our closed-loop agent workflow under deadline and got skeptical PMs to actually adopt it. Adoption is where most AI projects die — Nikhar made it the thing they couldn't work without.
First product hire who could translate VC pitches, HIPAA architecture, and pharmacy API integrations in the same week. Raised $250K and converted 4 clients in 3 months — that's a 0→1 operator.
Nikhar's MŌD client work synthesized 15+ SME interviews and Pinecone/Claude RAG prototypes into a board deck that informed $12M in cost savings. Strategic and technical in equal measure.
Paraphrased from performance reviews and reference calls. Full references available on request.
Traveling to new places and experimenting in the kitchen — two creative outlets that keep me curious.
Straight answers — fit, depth, timing, logistics.
Open to AI Product Manager roles at ambitious companies.
nj222@cornell.edu · +1 (607) 595-8222 · Irving, TX