← Back to projects
liveClient work·2026.07 – 2026.10
MyCREAM
An AI interactive-content platform cast with real influencers — I owned media, recommendations, tracking and localization
- Client
- Global consumer startup (joined as an embedded contractor)
- My role
- Media architecture, Recommendations, Pixel tracking, Localization
- Languages
- 10
The challenge
Product direction shifted constantly with retention and engagement data, and tickets were picked up the moment they landed. The job was to build things that would survive the next pivot without a rewrite.
Key decisions
One media layer for local, staging and production
I designed photo/video curation and delivery so local development, staging and production all share the same media assets, instead of maintaining media per environment.
Shipping through an AI coding agent
Almost all implementation went through Codex. I set structure and boundaries first so fast-changing tickets could ship quickly without the codebase sprawling.
Results
- Photo/video curation and media delivery
- Content recommendation algorithm
- Ad pixel tracking
- Localized UI and content into 10 languages, including RTL Arabic
Tech
- Media architecture — Photo/video curation; media shared across local, staging and production
- Recommendations — Content recommendations tuned to retention and engagement
- Pixel tracking — Pixel tracking for ad conversion
- Localization (i18n) — UI and content translated into 10 languages
- TypeScript
- Monorepo (Nx) — Turborepo + pnpm workspaces
Planning & collaboration
- AI-assisted build & review — Ticket-by-ticket delivery through Codex
- Small-team delivery — Embedded in a multi-dev team working straight off tickets