A product case study · fictional platform, synthetic data

I build AI and Data products that are intuitive, scalable, sustainable.

Two things live here. A walkthrough of PO Project Assignment: SeaTrace, a traceability product I owned end to end with a 14-person team. And a set of learning artifacts: "NewMed" is a fictional platform I built to learn US medical billing and to show how I work, from the first SQL query through to a working prototype.

SeaTrace: a product walkthrough See the learning artifacts
LinkedIn Download CV Contact
01 / Learning artifacts

How I ramp into a domain: by building in it.

These two are deliberate learning exercises, not products. Both run on synthetic data on a fictional platform I named NewMed. I built them to teach myself how US medical billing actually works: the claim lifecycle, why claims get denied, and where the manual effort goes. Everything below is something I can walk through and explain.

Flagship · Claims & billing AUC 0.853 · 80% preventable · $46k/mo protected

Denial Prevention Co-Pilot

A data exercise. I generated a synthetic warehouse of 6,000 claims, wrote the discovery SQL, and used it to learn the revenue cycle end to end. The data said roughly 80% of denials trace to causes that are knowable before a claim is submitted: eligibility gone stale, prior auth missing, a modifier absent, timely filing about to lapse. So I wrote an AI PRD and built a prototype that scores each claim before it goes out, explains why in biller language, predicts the likely denial code, and offers a one-click fix the biller can undo.

What I did: generated the dataset, wrote the discovery SQL, specified the model and how it should be evaluated, and built the interface.

Open the walkthrough
Companion · Practice operations 3,324 h/mo · 20.8 FTEs · 69% automatable

Practice Ops Bottleneck Analyzer

A process-flow exercise. At Deloitte I removed 100+ analyst hours a month by instrumenting a process first and automating second. This points the same method at practice operations: mine the task events, rank where the manual hours actually go, and pair every bottleneck with a specific automation and the hours it would return. It is the shape of the customised, data-grounded report I would want to produce for real practices once I understood their process and their data.

What I did: modelled the task-event data, wrote the analysis that ranks the bottlenecks, and built the dashboard and the adoption what-if.

Open the walkthrough
Real product · Owned end to end 14-person pod · 4 modules in ~3 months · 2-week sprints

SeaTrace: farm-to-fork batch traceability

The real one. A shrimp exporter ran on a legacy ERP that could not answer the only question that matters after something goes wrong: which pond, which day, which batch. I owned the backlog and ran the cadence for a 14-person team building the traceability platform that could. This walkthrough follows the product owner story end to end: the problem, the team, how I ran execution, what shipped, and what I would do differently.

Open the walkthrough

Behind the Co-Pilot there is also a written product requirements pack: functional specs, a technical requirements document structured for engineering, design requirements and an evaluation plan. Available on request.

02 / About

The rest of me.

I'm a product person who likes turning messy problems into things a team can actually use. Over the last five years I've worked across AI and data products, usually close to the delivery: writing the SQL myself, turning requirements into specs engineers can build from, and owning a backlog through to release. NewMed is what my ramp into a new domain looks like: from unfamiliar vocabulary to a working, evaluated product.

Outside of work, you can usually find me on the trails training for marathons, working with oils and charcoal, or looking at problems through a sustainability lens. I believe in giving back, whether that's sharing culture with kids or helping build digital literacy.

I try to stay honest about what I don't know yet, and I'd rather show working software than a deck. If any of this is useful to you, I'd be glad to talk.

LinkedIn Download CV Email me
PatternLab.AI · Product Strategy & GTM Lead
Nov 2025 to present · Hyderabad

Owned an MVP end to end with a 14-person cross-functional team: 4 modules shipped in 3 months, engagement up more than 75%, with A/B tests on UI and LLM prompts.

Deloitte USI · Consultant, Strategy & Analytics (AI & Data)
Aug 2023 to Nov 2025 · Hyderabad

Embedded with a large investment manager's data office: production dashboards, reporting automation saving 100+ analyst hours a month, and an ML churn model with a real-time scoring API.

M.Tech, IIT Delhi · B.Tech, DTU
Energy systems · DAAD Scholar at TU Munich

Studied energy and environmental systems; carried a habit of learning new domains by building in them.

Endurance running

Endurance running

Marathons and obstacle courses. Pacing and patience on the trails carries over into how I ship.

Painting and drawing

Painting & drawing

Oil, acrylic and charcoal. I led my college's art society and was a National Finalist at Red Bull Doodle Art.

Sustainability

Sustainability

It runs through my studies and much of my work; more a lens I carry than a hobby on the side.

Giving back

Giving back

Sharing culture with school children abroad, and volunteering on digital-literacy and education access wherever I land.