Headline

Product Manager · Applied AI & B2B SaaS @ NiryatAI | 0→1 platform builder, eval-first AI, RAG | Ex-Data Scientist | IIT Kanpur

About

I'm a product manager who builds 0→1 in ambiguous, high-stakes B2B problem spaces — and ships AI that has to be defensible, not just demoable. As the first PM at NiryatAI, I drove the platform build across buyer, supplier, and ops surfaces — system and workflow design, metrics framing, and a 6-initiative roadmap — turning a bespoke, trader-led business into a standardized product. My edge is applied AI, built to hold up in production. I shipped 7+ live AI systems across the procurement lifecycle — including AI-assisted supplier matching, RFQ extraction, document validation, and visual quality checks — each evaluated before launch and kept human-in-the-loop on high-stakes decisions. I scoped an 18-month roadmap and a post-order ops copilot. Outside work, I build to learn — LectureSnap is a solo-shipped AI product (extension, web app, API) that turns lecture videos into structured study notes, where I owned the whole stack, evals and all. It's the quickest way to see how I actually build; code and write-up are linked below. Portfolio: https://ds-v.github.io/portfolio/artifact-6-lecturesnap.html Code: github.com/DS-v/Lecture-Notes-AI

Location

Noida

AI stack

antigravity

codex

claudecode

Work experience

First Product Manager at NiryatAI, Dec 2023 - Present · 2 yrs 9 mos, Noida, First Product Manager at NiryatAI(formerly Worldref) driving the 0→1 build of a B2B cross-border procurement platform — buyer, supplier, and ops surfaces, RFQ through order execution. - Architected end-to-end system and workflow design, separating reversible decisions from irreversible commitments (approvals, PO issuance, pricing locks); structured around three cold-start bottlenecks — buyer reliability, supplier liquidity, ops leverage. - Designed AI-assisted supplier matching (precision over recall, offline-eval-tuned thresholds), lifting supplier response rates 15% → 35%; built supplier liquidity via capability-ingestion pipelines, growing the active base ~180 → ~1,400. - Shipped 7+ live AI systems across the lifecycle — including AI-assisted supplier matching,, RFQ extraction, document validation, HSN and hazardous classification, visual QC etc — with human review gated to high-consequence decisions. - Designed buyer-side PaaS monetization (rule-based fees + ML fulfilment-charge model), replacing ad-hoc deal-by-deal pricing. - Cut time-to-first-confident-quote 5 days → 2 and reduced post-PO escalation rate by 22%; built the metrics framework separating leading quote-quality signals from the RFQ→PO north-star. - Scoped an 18-month, 6-initiative roadmap; two AI-led — a post-order ops copilot (90+ user stories, per-action autonomy gradient, eval-gated) and a risk & delay-prediction system on a personally-built 120+ risk taxonomy. - Managed one APM — owned roadmap, user stories, and AI design decisions; front-loaded the "why" before handoff, raising autonomy and cutting rework.

Personal goal pursuit at Career Break, Nov 2021 - Dec 2023 · 2 yrs 2 mos, Prepared for UPSC Civil Services and Regulatory Bodies examinations

Data Scientist at Merilytics, Jun 2021 - Oct 2021 · 5 mos, Hyderabad, Built a hybrid ML + rule-based freight-routing model that cut logistics costs by ~8% across key lanes, and reduced high-density scheduling conflicts via Gantt/heatmap dashboards.

Education

Bachelor of Technology, Mechanical Engineering, Indian Institute of Technology, Kanpur