Ten years bridging platform teams and frontline users taught me one thing: the best PRD is worthless if it doesn't land on the ground. Now I write for that person first — with the PM craft to back it up.
Below: case studies, product skills, and the honest story of the transition. PRDs, prototypes, and evidence of thinking — no invented metrics.
Four end-to-end product cases — problem, research, prioritization, MVP, trade-offs, and what I'd do next. Expand any card to read the full write-up. Want the PRD or deck behind a case? Just reach out.
Budget gym beginners don't drop out because they don't know what to do — they drop out because deciding takes effort on a busy week. GymBuddy generates an adaptive weekly plan and a check-in simple enough to survive that week. Built, shipped, and tested with 10 real users.
GPS pickup friction at complex locations (malls, gated communities, airports) drops rider satisfaction and increases cancellations. Designed a suggested-pickup layer end-to-end.
Low AOV isn't a discount problem — it's a spend-anxiety problem. Segmented users, ranked interventions, designed the A/B test.
Retention doesn't come from streaks; it comes from someone noticing when you don't show up. Designed the accountability layer.
Additional discovery sprints and prototypes.
A timed, real-brief product analysis for a YC-backed AI healthcare company: I mapped the current user journey, pulled out where trust breaks down for a clinical user, sized the highest-leverage gaps against effort, and wrote up a prioritized set of recommendations with the metric each one should move. The work is shared as-is — walkthrough deck and analysis doc.
Open analysis folderIf Drive asks for access, the folder is still set to restricted. Please set it to "Anyone with the link can view" inside Google Drive.
Five JTBDs framing the 10-minute grocery user. Stack-ranked to three high-leverage bets: Smart Recurring Lists, Post-Order Cart Modification Window, and a Self-Serve Returns Tab.
PRD-first build: user flows, edge cases, and acceptance criteria defined before a single screen was designed.
Turns a pile of unstructured app reviews into ranked, actionable themes so a PM can see what to fix first instead of reading 1,000 comments.
Restaurants and event caterers throw away edible surplus because finding a taker in the last hour is manual work nobody has time for. I built an autonomous agent on Make.com that closes that gap: a donor submits a short Google Form (what food, how much, pickup window, location); the scenario fires on submission, calls an HTTP search step to find NGOs and food banks near that pin, and logs every request as a row in Google Sheets so there's a permanent audit trail. A Make AI Agent then reads the row, matches the surplus to the best-fit NGO on distance and pickup window, drafts the outreach message, and sends it — with the donor CC'd so the handover is verifiable. Product decisions: I kept humans in the loop for the final confirmation (an agent should not promise food it can't deliver), used a Sheet as the source of truth so anyone non-technical can audit or override, and scoped v1 to a single city radius rather than a national matching problem. Success measure: kilos rescued per week and time from submission to confirmed pickup.
Each case distilled to the one interaction that mattered. Phones marked Recording play a real screen capture of the build I shipped — no shareable link, so the work stays mine.
Grouped by what they're for. Not a checklist — the things I actually reach for.
A career changer's honest note — what I'm bringing over, and what's genuinely different about my lens.
For a decade I've watched platform decisions land — or fail to land — with the people actually using them. I've been the one running the UAT, training the rollout, and hearing the frontline groan when a feature misses the reality it's shipped into. I want to be upstream of that. Product is the seat where those calls get made, and I've spent ten years learning what makes them go right.
UAT rigour translates directly to airtight acceptance criteria. Running training rollouts taught me that adoption is a design problem, not a comms problem. Building a QA framework with measurable KPIs was, in effect, a north-star exercise. I've already done the unglamorous half of PM work — I'm here for the strategic half now.
Most PMs meet friction in a dashboard. I've met it in a support queue, in a training room, and in the escalations that never made it to Amplitude. That means I default to "will this actually get adopted?" before I ask "is this shippable?" — which turns out to be the question that saves teams from shipping the wrong thing.
Same format I use in PRDs. It reads honestly — no invented job titles, no team-of-30.
Through Accenture's community programme I taught students across multiple subjects in a village in Gadchiroli — one virtual room, the entire class at once, patchy connectivity and mixed comprehension levels. I front-loaded the one idea that mattered, swapped explanation for worked examples when the room went quiet, and checked understanding out loud. Designing for the person with the worst connection and the least context is exactly how I think about a first-run user.
The formal side of the transition — programmes, certifications, and what each one actually gave me.
The programme these case studies were built inside. PRDs, prototyping with AI tools, prioritization frameworks.
Foundations for AI-integrated product work — prompting, model choice, evaluation.
Sprint mechanics, backlog grooming, ceremonies — applied inside every case study.
Business management foundations.
Actively looking for APM / PM roles in B2C, mobility, fintech, and AI-first products. If your team wants someone who can write the PRD, run the rollout, and still talk to frontline users — I'd love to talk.
Curious how I'd handle a vague PRD, walk through a project, or stack-rank a roadmap? Tap the chat at the bottom-right. Trained on this portfolio, powered by Lovable AI.