Product Analyst @ HCL Healthcare · IIT Kharagpur '24 · Noida, India
Open to PM, APM and AI PM roles across consumer product, health and applied AI
I like figuring out why people don't use a product, and then fixing it.
I own engagement on a consumer health platform with 1M+ registered users: the roadmap, the PRDs, and what ships in what order.
Most of my work sits at the intersection of product, data and applied AI. Figuring out what problem is actually worth solving, writing the PRD, working with engineers, shipping it, then looking at the numbers afterwards.
I don't build features because they sound good.
The metric has to move. The user has to care.
Portfolio · Case studies · LinkedIn · Email
Sometimes the best product improvement isn't a new feature.
My roadmap said engagement. The app took fifteen seconds to open. Anything shipped onto that only ever gets measured on the people patient enough to wait, so I held my own roadmap for eight weeks and put launch time first.
- App launch time: 15s → under 2s
- Bundle size: 25MB → 6MB
- Step-sync completion: +35%, after the launch-time fix
Steps Premier League turned step tracking into a competitive season: eight teams of seven, five days, steps scored as cricket runs.
Session time moved from 3.5 → 7.8 minutes.
It is also a north star I would not choose again, because a product that works can shorten a session. The case study says so out loud.
See the live league → · Read the case study →
I took the AI Smart Health Report from 0 to 1: the product requirements, the UX flow, and the personalisation logic that decides what an individual report says.
It was designed to turn a page of clinical values into something a non-clinical adult can read, with one prompted next step.
The report was named as a differentiator in 5+ enterprise deals, and the cross-sell hooks I specified drove 15% incremental revenue.
I built the product. I didn't close the deals.
Where users drop off, why they don't come back, and what makes a health app worth reopening on a Tuesday.
- +20% daily active users in 12 weeks, from an engagement suite of challenges, streaks, live events and trackers
- +15% health assessment completion
Grounded is an evaluation harness that scores LLM-generated health summaries against a rubric you can read: grounding, scope, escalation, readability. Every check is a deterministic rule, so it runs in your browser with no model call and no API key.
Its first run disagreed with my own labels on 5 of 16 cases. Most of those were the harness being wrong rather than the labels, and some were my labels moving toward the harness, which is exactly the failure mode an eval set is supposed to prevent. Both versions are still in the repo so a change can be diffed.
Problem → Evidence → Hypothesis → Build → Measure → Iterate
Before building anything, I try to answer:
- What problem are we actually solving?
- How do we know it's a real problem?
- Which users hit it hardest?
- What's the smallest thing that tests the hypothesis?
- What metric should move if we're right?
- What did we learn after shipping?
I don't treat AI as a separate "AI project". It's how I work.
ChatGPT · Claude · Cursor · Claude Code for PRD and spec drafting, user research synthesis, SQL exploration, edge-case discovery, experiment ideation, data analysis, and building small tools before involving engineering.
The goal isn't to replace thinking. It's to get from idea to testable output much faster.
Product
Product Discovery PRD Writing User Stories A/B Testing Roadmap Planning 0→1 Launches Agile/Scrum Stakeholder Management Metrics/OKRs
Data
SQL Python Mixpanel Amplitude Tableau CleverTap Google Analytics
AI
LLM APIs Prompt Engineering AI Personalisation Rapid Prototyping
Tools
JIRA Figma Confluence Notion
HCL Healthcare · Product Analyst · Oct 2024 to present
Engagement on a consumer health platform with 1M+ registered users. The four cases above, plus health assessment journeys, cross-sell, analytics and experimentation.
Circle Health · Product Intern · Jul to Sep 2024
Owned discovery on the claims journey, mapped the drop-offs, redesigned the flow, and built a real-time dashboard tracking 50K+ user journeys.
Droom · Product Management Intern · May to Jul 2023
Listing fraud detection validated by A/B testing, and a vehicle valuation page rebuild that grew report purchases 16%.
Infinyte Club · Product Operations · Feb to Apr 2023
Found a five-step verification wall killing signups and proposed deferring it past the signup boundary. Completion doubled.
YourStory Media · Product Management Intern · Jun to Sep 2022
Discovery across 5,000+ surveyed users, and the go/no-go recommendation that shaped a new vertical.
IIT Kharagpur · Dual degree · 2019 to 2024
A number without its population is a claim. A number with one is evidence.
A dashboard doesn't tell you what to build. It tells you what to investigate.
AI makes good PMs faster. It doesn't make bad product thinking good.
I'm looking for roles where I can take more ownership of product problems, work closely with engineering, and build products that reach a lot of people.
If you're building something interesting, especially around Consumer Tech, Healthcare, AI or Growth, I'd love to hear from you.
