The best AI PMs I know spend 80% of their time on evals and 20% on prompts. Here's the loop I run every week…
AI ProductManager.
Building and shipping real products before my first PM role.
About
I've built and shipped four products end-to-end — Kartify, CafeOS, Tapinfi and FinMate — before ever holding a Product Manager title. Each one ran through the full product loop: customer discovery, JTBD framing, prioritization, scoping, shipping, and post-launch measurement. The craft is already the work I do.
My BA and QA background is not a side story — it's the same PM work under a different job title. Requirements gathering, acceptance criteria, stakeholder negotiation, bug triage and release validation are exactly what Product Managers do before a feature gets written. The self-initiated products are even stronger evidence: nobody assigned them, nobody paid me for them, and I still chose the problems, ran the research and shipped the builds. That kind of unprompted judgment is harder to teach than a title.
I want to be one of the leading AI Product Managers of the next decade. Not for the title — because the products that matter will be the ones that explain their reasoning, not just their output, and that's the same principle I designed into Kartify. That's the trajectory I'm on.
Shipped Products
Shipped and prototyped from 0 to 1 — click any card to inspect.
Case Studies
Problem → Research → JTBD → PRD → Metrics → Lessons.
Shoppers waste hours comparing products across marketplaces with inconsistent specs, reviews and pricing.
12 user interviews across three shopping personas; competitive teardown of Amazon, Perplexity Shopping and Google Shopping.
When I'm buying a considered product, I want a trusted advisor that asks the right questions, so I can decide confidently without opening 20 tabs.
Conversational search with clarifying follow-ups, memory of preferences, structured comparison and a recommendation with reasoning.
Career Trajectory
Greenfinch Global Consultancy
Business Analyst & QA Intern
Gathered requirements, wrote BRDs and PRDs, ran QA cycles and validated user flows alongside developers and clients.
- ·Owned requirement clarity across sprints
- ·Bug triage and regression coverage
- ·Client discussions and acceptance criteria
Assert InfoTech
Business Analyst
Stakeholder communication, workflow analysis and functional specifications across product discussions and testing support.
- ·Requirement gathering and documentation
- ·Feature validation and acceptance criteria
- ·Process improvement across teams
Founder Track
Tapinfi
Built a SaaS platform enabling professionals to instantly share digital profiles using NFC-enabled smart cards.
Bharat Svarga
AI-powered travel platform focused on spiritual and heritage circuits in Rajasthan — starting with Jaipur–Pushkar–Ajmer — that pairs personalised itineraries with a curated network of local vendors (homestays, guides, transport). Onboarded ~25 pilot vendors in Jaipur before scaling into a second circuit.
Notes on how I build, decide and ship
Essays on discovery, AI products, growth loops and decision-making.
Customer discovery without a research team: 5 interviews, one spreadsheet, and the JTBD you actually ship against.
How I ship an MVP in 6 weeks — the exact week-by-week breakdown for Kartify.
Framework Library
The mental models I lean on to decide what to build, what to cut and what to test next.
Understand the underlying job customers hire your product to do — beyond features.
The Lab
AI experiments, prototypes and workflow tools I build for myself.
Prompt Engineering
Structured prompt libraries, evals and prompt-as-spec workflows.
AI Agents
Tool-using agents with planning, memory and reflection loops.
Multi-Agent Systems
Coordinated agents for research, comparison and synthesis tasks.
MCP
Model Context Protocol tooling to give assistants safe access to real systems.
n8n Automation
Product-ops workflows: research, outreach, monitoring and triage.
RAG + Vector DBs
Grounding LLMs in private knowledge with retrieval-quality evals.