AI Product Manager · Growth · Mobile
Senior PM with 10+ years building products that grow. She doesn't just talk about AI — she ships with it. Currently building Digital Nest using Next.js, Claude Code and Vercel.
A selection of product initiatives across two mobile apps — each with different constraints, the same approach: research, experiment, measure, iterate.
Japan was one of the strongest markets globally for the product category, yet activation, conversion and recurring subscriptions were consistently underperforming. QA and localisation audits found no technical issues — the problem had to be elsewhere.
Partnered with SEM and requested a market audit from Google to understand Japanese consumer psychology and purchasing behaviour. Combined those insights with behavioural analytics to identify hypotheses around pricing presentation, onboarding emotional connection and visual identity.
A roadmap of experiments was designed to validate each hypothesis: localised onboarding, redesigned pricing page, different price points, weekly/monthly/yearly plans, free trials and multiple visual identities. One experiment stood out: when shortlisting onboarding images, a Japanese colleague confirmed that three of the proposed models were actually Korean. Her recommended image — truly representative of Japanese users — became the winning variant, and the fully localised paywall outperformed all others.
The findings extended beyond this app. Japan was subsequently treated as a standalone market across the majority of the portfolio, with personalised approaches delivering results across multiple products.
The biggest optimisation opportunity wasn't technical — it came from understanding users in their local context.
Took ownership of a mobile scanning app that had spent over a year without a dedicated PM — almost no handover, limited documentation, a confusing backlog and significant functional gaps vs competitors.
Before proposing anything new, created visibility: built a realistic roadmap, benchmarked competitors, identified critical gaps and reconstructed the analytics layer from scratch — auditing events, removing duplicates, creating a consistent taxonomy and documenting the full measurement framework.
When a major OCR-powered search feature required a larger architectural refactor than expected — and an engineer went on long-term sick leave — reorganised releases so the OCR project continued in parallel while A/B tests and quick wins kept shipping continuously.
Experimentation velocity doesn't require perfect conditions — it requires good prioritisation and the discipline to keep shipping while solving harder problems in parallel.
Camera Translation — one of the app's most used features — had evolved through years of rapid experimentation into a fragmented experience. Two critical pain points emerged from research: users had to manually select the source language before scanning, and the translation frame required constant manual adjustment.
Combined behavioural analytics, user reviews, UX research and customer feedback to define the redesign scope. Migrated to a more capable OCR provider with automatic language detection, expanded language support and a cleaner translation overlay directly on top of the original text — removing friction from the core user journey without disrupting the experimentation cadence.
As products mature, sustainable growth increasingly comes from improving the quality of core experiences rather than continuously adding new functionality.
Following the Japanese market research, South Korea had been treated as a generic Asian market. Deeper analysis suggested user behaviour was actually closer to Western markets — a hypothesis worth testing rather than assuming.
Adapted acquisition and monetisation experiments to reflect Western market dynamics rather than Asian generalisations. Tested pricing, onboarding and messaging strategies aligned with the hypothesis that Korean users behaved differently from the Japanese market.
Those gains were later reduced following a marketing budget reallocation towards higher-performing regions — a business decision outside product scope. The experiments, however, successfully validated that market assumptions should always be tested, not inherited.
Generalising user behaviour across countries is a product risk. The value of this initiative wasn't only the revenue growth — it was proving that assumptions need data to survive.
Open to Senior PM and AI PM roles across Europe and remote-global markets.