Context-aware mobile apps: sensors, AI and social analytics
Apps that use smartphone sensors to understand the user’s context and offer the right content at the right moment, plus a dashboard that combines real usage with social trends to guide product decisions.
Mobile, AI and data
- Method
- Market research and tests with real users
- Privacy
- On-device processing where possible
- Data
- Product decisions based on evidence
The challenge
Content apps demand too much attention: random notifications, endless screens, generic suggestions. The goal was an experience that knows when to speak and when to stay quiet, and a way to learn what people actually care about before producing content.
How we worked
- Market research on competing products and user reviews, to find the most common reasons people drop off.
- Prototypes tested in the field with small groups of users, observing real use instead of relying on surveys.
- Product metrics defined before launch (activation, return visits, completed content) and reviewed every week.
- A roadmap updated from the data: little-used features get simplified or removed.
Architecture and choices
- Sensor fusion (location, motion, activity) to estimate context: walking, stopping, being near a point of interest.
- Lightweight on-device models for frequent decisions, cloud only for content generation: less battery, less latency, less personal data moving around.
- A voice-first, low-screen interface, with the AI asking permission before it steps in.
- A social analytics pipeline on public, aggregated data: rising topics, sentiment and seasonality feed the editorial plan.
External services, and why these
App
React Native with native modules
One codebase for iOS and Android, with Swift and Kotlin modules where sensors and low-latency audio need them.
Chosen over: Two separate native apps, Flutter
Product analytics
PostHog
Funnels, sessions and feature flags in one tool, hosted in the EU.
Chosen over: Google Analytics, Mixpanel
Social data
Official APIs and aggregated datasets
Public data obtained compliantly and stable over time.
Chosen over: Scraping
Outcome
Product in public beta, with an improvement loop driven by usage data and external trends rather than internal opinions.
Stack
- React Native
- Swift
- Kotlin
- Core Motion
- On-device ML
- LLM
- Python
- PostHog
- Mapbox