AIProductManagerwithanengineer'shands.
AI Product Manager with an engineer's hands. I spent 2+ years building fullstack products for millions of users, then moved toward deciding what to build. Now doing a Master's in AI at UNSW and building AI products end to end.
Where I've built.
Product Manager
May 2025 – Jan 2026- Took a 0→1 product from discovery to a production launch, owning the roadmap end to end.
- Drove discovery into PRDs, Figma flows, and an OKR-aligned roadmap; prioritised the backlog toward the MVP.
- Led a six-person engineering team, ran Agile sprints in Linear, and owned final PR review and prod merges. Cut cycle time ~20%.
- Ran competitor tear-downs and lean experiments (smoke, concierge) to de-risk bets before building.
Founder Fellow
Jun 2025 – Aug 2025- Full-scholarship fellow. A two-week sprint from idea and team formation to MVP, customer interviews, and investor-style pitching with 1:1 coaching.
Swim Teacher
Oct 2024 – Present- Teach children aged 3 to 10. Keep classes safe and structured. An ongoing lesson in patience, communication, and earning trust.
Software Engineer Intern
Jan 2025 – Feb 2025- Built 0→1 site components in Next.js and Convex; partnered with the CTO on the near-term roadmap and developer workflows.
Full Stack Engineer
Apr 2023 – Jul 2023- Built React/Gatsby UIs and Node/Sequelize REST APIs for an API-first geospatial climate-data product.
- Used MapLibre and D3 to make satellite datasets usable in client workflows.
Senior Software Engineer (SDE1)
Jan 2021 – Mar 2022- Shipped the listing, product-detail, and homepage journeys used by 50M+ recurring users.
- Led frontend for gift-card redemption; supported a festive spike that lifted revenue 65%.
- Cut JS bundle 20% and raised Lighthouse Performance 15% via code-splitting, memoization, and route-level lazy loading.
- Led Localization (Shopping in Hindi). Won 'Fastest Rookie to Rockstar'.
Freelance Consultant
Mar 2022 – Jan 2024- Automated enrolment, credentialing, and reporting with Moodle plugins, themes, and APIs. +30% engagement, -40% grading time.
- Led tech strategy and mentored a team of 5 interns; +25% efficiency with JIRA.
AI Intern
May 2018 – Jul 2018- Trained a DenseNet to 79% on CIFAR-10; prototyped behavior cloning for steering-angle prediction; explored Winograd convolutions with skip connections.
Foundations.
AI, deep learning, big data, NLP, and AI product design.
- Reforge — Product Foundations (full feature lifecycle: opportunity, design, development, launch)
- Machine Learning, Andrew Ng (Coursera)
- UNSW Peter Farrell Cup 2025, Startup Pitch (Semi-Finalist)
- Leadership Foundations (UNSW) + Strategic Thinking (AGSM)
- Swim Australia Teacher · CPR/AED/First Aid
- SSI Rescue Scuba Diver
What I work with.
What I learned at Reforge.
Reforge runs the programs that senior product leaders at companies like Meta, Stripe, and HubSpot use to level up. Product Foundations walks the full feature lifecycle, opportunity to launch, the way strong product teams actually run it.
Score an opportunity on user value, business value, and strategic fit before building. Size the upside with a funnel model and pitch it to leadership.
Run focused brainstorming against desirability, feasibility, and viability to land on the right solution, not just the first one.
Map stakeholders, set milestones, and manage risk through the sprint with defensive and offensive plays. Know the PM's job at each step.
Coordinate the launch, define how performance is measured, and run post-launch communication so the work actually lands.
Bullet-ready, if you're skimming.
The four AI products on this site, distilled into CV lines.
- • Wrote the PRD, metric framework, and experiment plan for each of four AI products.
- • Reframed model evaluation as a measured release gate, turning a vibes call into a defensible one.
- • Designed evidence-first and trust-by-design AI UX (citations, confidence, approval queues, audit) where the guardrail metric protects the user.
- • Scoped MVPs with explicit in/out calls and a credible path from MVP to V2.
- • Built and shipped four hosted AI apps (Next.js + TypeScript, Vercel) across evaluation, RAG, multimodal, and agents, each mock-first so the demos run with zero API keys.
- • Built an LLM evaluation harness: deterministic checks plus an LLM judge with per-run cost and latency, surfaced as a ship/hold release gate.
- • Built a RAG pipeline (chunk to embed to cluster to retrieve to generate) over user feedback with pgvector and a 2-plus-citation grounding rule.
- • Built a human-in-the-loop agent with a manual tool-calling loop, approval gates, and an append-only audit trail.
- • Interpretability research (UNSW team project): probed the residual stream of five language models with TransformerLens to detect input corruption the output hid; 0.96+ linear-probe accuracy, confirmed causal with activation patching.