← All selected work

Private, personalized mental wellness

Resilient Moment

AI support that did not require surrendering private thoughts.

The problem

Mental wellness tools can become most useful at the moment people are least comfortable sharing personal data. We wanted personalization without turning private reflection into a server-side profile.

The work

What I carried.

The Joyment app delivered personalized exercises informed by CBT, with sensitive user data processed and stored on the device.

  1. 01

    Owned the product and technical roadmaps and led 17 part-time contributors across engineering, product, design, mental health, and go-to-market work.

  2. 02

    Built the React Native app around on-device data and privacy-first defaults, then evaluated LLM, speech, voice, and image services for quality and cost.

  3. 03

    Ran a year-long beta with more than 130 users and input from more than 40 mental health professionals.

Selected outcomes

34languages
350+organic downloads
97%lower voice cost

Decisions that mattered

Put privacy in the architecture

User history and personal material stayed on the device. Privacy was a data-flow decision, not a paragraph added before launch.

Measure the whole experience

The team tested the exercise, language, voice, onboarding, and perceived value with users and mental health professionals throughout the beta.

Joyment showed me that useful AI depends as much on restraint and trust as it does on model quality.