AI Health Recommendations for Mobile Apps

AI Health Recommendations Based on Mobile Health Data

Most health apps show users numbers—and they close the app without changing habits. We change the scenario: instead of "here's your data" we give "here's what this data means for you today." According to HealthKit, the average user's daily activity is 5,423 steps, sleep—7.1 hours. Based on this data, our rule engine generates 12–20 candidate recommendations per day, and ML ranking selects 3–5 most relevant ones. For example, if resting heart rate is above 80 bpm and sleep is below 6.5 hours, the recommendation is to reduce workout intensity.

In this article, we explain how we build a recommendation system: from data collection to push notification delivery. You'll learn why a rule engine is not a relic but a foundation, and how ML adds personalization without losing transparency. We have developed 15+ health apps and have extensive experience; one of our projects reached 500,000+ installs. If you need a similar system, contact us—we'll help design and implement it.

Why a Rule Engine Is the Foundation of Personalized Recommendations

Rule-based approach is not outdated—it's practical. Rules are transparent, testable, and require no dataset for training. ML on top of rules adds personalization in ranking. Combined conditions matter: a resting heart rate of 85 bpm alone may be normal, but combined with sleep deprivation, it's a marker of overtraining.

Recommendation System Architecture

Personalized health recommendations are a pipeline with multiple layers, not a single algorithm. The table below outlines the main components.

Data: HealthKit as the Single Point on iOS

class HealthDataAggregator {
    private let store = HKHealthStore()

    func weeklyStats() async throws -> HealthWeekSnapshot {
        async let steps = fetchSum(.stepCount, days: 7)
        async let sleepHours = fetchCategorySamples(.sleepAnalysis, days: 7)
        async let restingHR = fetchAverage(.restingHeartRate, days: 7)
        async let activeEnergy = fetchSum(.activeEnergyBurned, days: 7)

        return try await HealthWeekSnapshot(
            avgDailySteps: steps / 7,
            avgSleepHours: sleepHours,
            avgRestingHR: restingHR,
            totalActiveKcal: activeEnergy
        )
    }
}

On Android, we use Health Connect SDK with HealthConnectClient.readRecords(StepsRecord::class) to query steps for a period. We also connect third-party devices via Core Bluetooth on iOS and BLE on Android to get data from scales, blood pressure monitors, and fitness bands.

How ML Improves Personalization

The rule engine generates a list of recommendation candidates. The ML model ranks them by the probability of the user taking action. We use gradient boosting on features: historical CTR, day-of-week pattern, streak adherence. We train on implicit feedback: shown → opened → completed (data from HealthKit).

How Push Notification Timing Is Determined

The right moment matters more than content. "Go to bed earlier" at 20:00 works; at 23:30 it doesn't.

func scheduleRecommendation(_ rec: Recommendation) {
    let content = UNMutableNotificationContent()
    content.title = rec.title
    content.body = rec.shortBody
    content.sound = .default

    let bestTime = optimalDeliveryTime(for: rec, userSchedule: userProfile.typicalSchedule)
    let trigger = UNCalendarNotificationTrigger(
        dateMatching: Calendar.current.dateComponents([.hour, .minute], from: bestTime),
        repeats: false
    )
    let request = UNNotificationRequest(identifier: rec.id, content: content, trigger: trigger)
    UNUserNotificationCenter.current().add(request)
}

optimalDeliveryTime analyzes user patterns: app opening time, sleep, workouts. Context binding: if CMMotionActivityManager shows the user is walking, don't show an activity recommendation.

Principles of Effective Recommendations

One specific recommendation per day beats five generic ones. "Walk 2,000 more steps by 6 PM, you're at 1,200 now" works. "Move more" does not. Context awareness and personalization are key drivers of engagement. We use A/B testing to select the most effective formats and content.

Work Process for Recommendations

1. Analyze data sources and design HealthKit/Health Connect schema.
2. Develop 20–50 rules based on medical protocols and typical scenarios.
3. Implement ML ranking using synthetic data for training.
4. Build delivery system: in-app widgets + push with optimal timing.
5. Run A/B tests for at least 2 weeks.
6. Document API and train the client's team on rule administration.

What's Included in Our Work

• API documentation for integration and rule management.
• Training the client's team on administering and customizing recommendations.
• Post-launch support for 3 months.
• Code warranty: fixing bugs within the specified architecture.

Timeline Estimates

Exact estimation depends on integration complexity and number of data sources. Contact us for a consultation—we'll analyze your project in 1 day. Get an individual proposal.

Machine Learning in Mobile Apps: CoreML, TFLite, and On-Device Models

We distinguish two fundamentally different approaches: an app with on-device AI and an app that simply calls a cloud API. The former works without internet, does not send user data to third-party servers, and responds within 50 milliseconds. The latter depends on network latency and pricing plans. Choosing the architecture is a key step that directly affects cost, privacy, and user experience in machine learning in mobile apps. Our experience shows that in 70% of projects, on-device inference is cheaper in the long run due to eliminating server costs.

How to Choose Between CoreML and TFLite for On-Device Inference?

CoreML — Apple's native framework for running ML models on device. Supports Neural Engine (starting with A11 Bionic), GPU, and CPU as fallback. Models are converted to .mlmodel format via coremltools from PyTorch, ONNX, or TensorFlow. Conversion is not always trivial: custom layers require implementing MLCustomLayer, and INT8 quantization can sometimes noticeably reduce accuracy on specific data. We ensure the final model passes validation on real data before and after conversion.

TensorFlow Lite — cross-platform alternative for Android and Flutter. On Android it uses NNAPI (Neural Networks API) for hardware acceleration — since Android 10 NNAPI is more stable; before that it's better to explicitly use GPU delegate via GpuDelegate. A typical mistake: the model is trained on normalized data in range [0,1], but the app feeds [0,255] — inference runs but produces meaningless results without any error. We include an automatic input data validation module in the SDK.

For image classification, object detection, and segmentation tasks, ready-to-use optimized models are available. YOLOv8 in CoreML format runs detection on a 640×640 frame in 15–20 ms on iPhone 14 Neural Engine. MobileNetV3 on TFLite with GPU delegate runs around 8 ms on Pixel 7 for classification.

What If You Need Text Generation On-Device?

Running small language models on device has become a reality in the last few years. Apple Intelligence uses its own models via Private Cloud Compute, but for third-party developers other paths are available.

llama.cpp with Metal backend on iOS is a working approach for phi-3-mini (3.8B parameters, 4-bit quantization, ~2.3 GB). Inference: 15–25 tokens/second on iPhone 15 Pro. For integration in Swift, use the Swift Package llama.swift or a wrapper via C interface llama.h. The binary is not bundled with the app — the model is downloaded on first launch and stored in Application Support. Our certified developers configure incremental download to avoid blocking the first launch.

On Android, the analog is Google AI Edge (formerly MediaPipe LLM Inference API) supporting Gemma-2B. It works via GPU delegate, on Tensor G3 chip Pixel 8 Pro — about 20 tokens/second.

Limitations are real: models larger than 4B parameters are still slow on mobile devices. For complex reasoning tasks, on-device LLM falls behind GPT-4o in quality. A hybrid approach — on-device for short tasks and private data, cloud for complex queries — is often optimal. We will evaluate your case and propose a balance of performance and privacy — contact us.

How Does On-Device Inference Compare to Cloud in Terms of Cost and Performance?

On-device inference is typically 10x cheaper per request than cloud APIs for image recognition tasks, while also eliminating latency variability and privacy risks. The table below summarizes the trade-offs.

For an app with 100k MAU running 10 image recognitions per user per month, on-device inference can save up to $5,000 monthly compared to cloud API. Get a free consultation on your ML architecture today.

Integrating OpenAI API and Other Cloud Models

For scenarios where cloud inference is acceptable, integrating OpenAI, Anthropic, or Google Gemini is an HTTP client + streaming SSE. In Swift, AsyncThrowingStream is convenient for streaming responses. In Kotlin, use Flow.

Critically: API keys must never be stored in the app bundle. Even an obfuscated key can be extracted from the IPA in 10 minutes using strings or frida. Correct architecture: mobile app → your own backend → OpenAI API. The backend controls rate limiting, logs requests, and protects the key.

What Is Included in the Work (Deliverables)

• Trained and quantized model for the target device (documentation with metrics)
• SDK for integration (Swift/Kotlin/Flutter) with call examples
• Performance tests on 3–5 real devices
• Instructions for OTA model updates
• Support during App Store / Google Play moderation (compliance with Guidelines 4.2, 5.1)
• 2 weeks of technical support after release

Typical Project Pipeline

1. Task analysis — measure latency, privacy, size, supported devices.
2. Model prototyping — in Python, evaluate accuracy on target data.
3. Conversion and quantization — for CoreML/TFLite with validation.
4. Integration into the app — model wrapped in a service layer (easy to swap CoreML ↔ TFLite ↔ cloud).
5. Testing — on real devices, measure FPS, RAM, battery.
6. Deployment — via TestFlight / Firebase App Distribution, monitor metrics.

Timelines: integration of a ready CoreML/TFLite model — 1–2 weeks, development of a custom model with mobile optimization — from 6 weeks, on-device LLM chat with personalization — 4–8 weeks.

Why We Take on Complex Cases?

10+ years of experience in mobile development, 50+ implemented AI/ML solutions, guarantee of compatibility with current iOS and Android versions. All projects undergo code review and load testing. The cost includes preparation of moderation documentation and training of your team.

Contact us — we will help you choose the architecture and implement ML in your app turnkey. Order an audit of your existing solution — we will assess the potential for server cost savings free of charge. In some projects, savings can reach significant amounts per month.