About
8 years in product, 14 in tech. I build AI in production to automate ops-heavy B2B - distributor onboarding, logistics, ERP integrations, payments - and to accelerate delivery through agentic tooling.
My career has one repeating shape: do the work by hand long enough to see the structural problem, then build the system that removes it.
I don't optimize individual steps. I look for the one constraint slowing everything down, redesign the system around it, and use AI where it creates real leverage - not as a chatbot bolt-on, but as something that does actual work for real teams. The result is a Builder PM who ships AI features in production and writes code alongside engineering, owning the full lifecycle from prompt design to deployed code to adoption measured in PostHog.
The pattern began in operations. I spent five years on CAKE POS - Sysco's restaurant point-of-sale and payments product - progressing from Application Support Engineer to Operations Lead on the payment gateway, building the support team from scratch. Living in the support data taught me where the product actually broke. When I moved into the Proxy Product Owner role (2017–2019), I owned the payment gateway end-to-end: I doubled users, grew revenue 120% ($40M), and traced a broken chargeback investigation workflow back to its root cause, rebuilding the tool with engineering to save $50,000/month. Seven years in payments and POS is a foundation that still differentiates me for B2B fintech.
At Cut+Dry - a Series B B2B SaaS company digitizing US foodservice distribution - I went deep into operationally complex product. I built Track, a 0-to-1 last-mile delivery product, validating demand with a third-party integration before building, then killing a live-map feature no one used and reframing the whole product around ETA notifications. It launched across 4 distributors and 8,000+ restaurants, including a 2,500-restaurant enterprise rollout at midnight-departure scale.
Then I moved into onboarding - and this is where the operator-then-builder move played out fully. I onboarded 29 distributors directly, running nightly calls with US partners from Sri Lanka while working with engineering during the day, integrating across every major foodservice ERP. I trained an analyst who independently onboarded 5 more. Across both roles: $72.7K/month in MRR across 34 accounts at 94% retention, including all four regional rollouts of a $1B+ distributor that drove $5.32M in new online revenue.
A note on titles: "Senior PM, Onboarding" can read like program management. It wasn't. I built features to unblock launches - pricing-control overrides, customer-editable settings lockdowns, dual-source inventory logic - and each became a reusable platform capability. The PM who does the onboarding sees product gaps the PM who only reads tickets never will.
After 34 manual onboardings, I'd seen the same bottlenecks on every account. I vibe-coded a prototype to earn the mandate, then built an AI-enhanced onboarding platform from scratch with one engineer - cutting time-to-revenue 40% (189 to 113 days) and requirements capture 4x (100 to 26 days). I PM'd 5 AI features into the product (discovery, requirements, testing); on three I also shipped the prompt design and user-side behavior end-to-end: Ask FDE (a RAG-powered Q&A assistant), Jira ticket creation, and a weekly-update drafter. Engineering owned the build on the other two (meeting matching, spec generation).
The part I care most about is whether real users adopt what I ship. Every AI feature I shipped is instrumented in PostHog, and the combined usage grew ~51x in 6 months - from 7 events in January 2026 to 357 by 06/26/26, with 32 unique monthly analysts at peak. All three are now part of the analyst team's daily workflow. That's the distinction I'd draw: an AI PM isn't someone who specs AI features. It's someone who can tell you, with evidence, which ones real users actually reach for, and prioritize from there.
Most recently I wrapped delivery itself in a 5-agent pipeline - discovery, prototyping, PRD, testing, orchestration - building the orchestration agents around an autonomous coding engine while keeping me at the review gate. I don't just ship features; I build the machine that ships them.
The onboarding platform started as a v0 prototype. I built a working version with a React frontend, Node.js backend, and MySQL database using Cursor - enough to earn the mandate to lead product for onboarding. We scrapped that prototype and built the real platform from scratch with one engineer.
For the first few weeks I only vibe-coded frontend changes; the engineer owned backend. Within 3 months we'd built the harness and hooked up the agents that let me work end-to-end - both frontend and backend, on real production code. Now I ship entire features myself: spec, frontend, backend, deployment. Engineering reviews at the gate instead of building from spec.
The 74 PRs and 44K lines on the homepage aren't a side project. They're the result of moving the constraint: from "I write specs and wait" to "I ship the feature and someone reviews it."
An AI PM or Senior PM role at a tech or fintech company in Kuala Lumpur or the wider SEA region - somewhere with AI in production today, genuine product ownership from discovery through delivery, and operationally complex problems worth solving. I'm based in Colombo, Sri Lanka and relocating to Kuala Lumpur. The domains I've lived in - dealer/partner onboarding, last-mile logistics, AI features with measured adoption, and payments - map directly onto the companies building in SEA right now.
I ran a 4-hour hands-on workshop, "Beyond Vibe Coding," for CS students at the University of Moratuwa (Sri Lanka's #1 ranked CS program). The core thesis: shift from vibe coding (prompt, accept, move on) to agentic coding by building the framework that makes agents useful and shipping repeatable. Most students shipped a real feature by session end.