Case study 03 · Digital health · 8 months
A diet plan for one person at a time, 94% accurate.
BetterHealth helps people reach wellness goals through personalised diet plans and fitness tracking. Algobain built the recommendation engine, the cross-platform app and the cloud backend beneath both.
One-size-fits-all plans did not fit anyone.
BetterHealth set out to combine nutritional science with software that adapts to each user. What stood in the way:
- iUsers struggled to find diet plans that fit their lifestyle and health goals.
- iiNutrition tracking and fitness activity lived in separate places.
- iiiGeneric plans made it hard to stay consistent.
- ivThere was no real-time feedback on nutritional intake or exercise progress.
A model at the centre, not bolted on.
AI personalisation engine
A machine-learning model that generates a diet and fitness plan for each user.
- A user profiling system capturing health metrics, preferences and goals.
- A recommendation algorithm analysing 50+ parameters, including BMI, activity level and dietary restrictions.
- Dynamic meal planning across 1,000+ recipes with nutritional data.
- Real-time adjustment based on each user's progress and feedback.
Cross-platform mobile app
One codebase for iOS and Android, built for daily use.
- Built in React Native.
- Offline-first, so tracking never stops for a lost connection.
- Automatic activity sync with HealthKit on iOS and Google Fit on Android.
- An interface designed around daily engagement and habit formation.
Backend infrastructure
A cloud backend that processes user data as it arrives.
- REST APIs in Node.js and Express.
- MongoDB for flexible storage and fast queries.
- AWS infrastructure with auto-scaling.
- Real-time notifications for reminders and achievements.
- Secure authentication with JWT and OAuth.
03 · Architecture
Every meal logged makes the next plan better.
The profile, the wearable and the user's own feedback feed one model. The plan it generates goes to the phone, and what the user does with it comes straight back.
Fig. 1 Reference architecture as built. Component names are generic.
The plan is the product. It changes when the watch syncs or a meal is logged, and tells the user why.
The numbers the platform is held to.
| Measure | Result | Basis |
|---|---|---|
| Operational cost saved | $180K/yr | [Cost baseline] |
| User retention | 78% | [Cohort and period] |
| Recommendation accuracy | 94% | [Evaluation method] |
| App rating | 4.7/5 | [Store and date] |
Figures from the engagement record. Where the basis of a figure is not yet documented, it is shown as a placeholder rather than invented.
What it runs on.
- Mobile
- React NativeHealthKitGoogle Fit
- Services
- Node.jsExpressJWT, OAuth
- Data
- MongoDB
- Model
- ML recommendation engine
- Cloud
- AWSAuto-scaling
A product that should be built around a model?
Tell us what you are trying to run in production. A senior engineer replies within 48 hours.
- 1Within 48 hoursA senior engineer replies, not a sales desk.
- 2NDA, then a working sessionWe sign first, then look at the system together.
- 3Two-week fixed-fee assessmentWritten roadmap, ROI and build-versus-buy. No lock-in.