Applications

WTMF App

2024App Store

DeepAI Integration

iOSPlatform

<500msLatency

About WTMF App

WTMF (What The Mind Finds) is an AI-driven mobile application that delivers personalized, intelligent experiences — bridging the gap between complex AI capabilities and everyday user accessibility.

Built with Flutter (Frontend) and a Python-powered AI Backend, WTMF brings on-device and cloud AI inference together in a seamless mobile experience.

🤖 AI-powered personalization
⚡ Sub-500ms response latency
🧠 On-device & cloud hybrid inference
📱 Native iOS experience

Vision: To democratize AI — making powerful, intelligent tools feel natural, fast, and effortless for every user.

How It Works

AI Inference Engine

Hybrid On-Device & Cloud Inference

Lightweight AI models run directly on-device using edge inference frameworks, handling fast, low-latency tasks with no network dependency. Heavier, compute-intensive requests are offloaded to a scalable Python cloud backend via optimized API calls.

Flutter (Frontend)Python AI BackendEdge AI (on-device inference)FastAPI / REST

Personalization Layer

The app learns from user interactions and preferences to tailor AI responses over time. A lightweight user-context model is maintained client-side and synced with the backend to improve response relevance.

Firebase Firestore (user profiles)Python ML modelsRiverpod (state management)

Backend Architecture

Python-Powered AI Services

All heavy AI computation runs on a Python backend — model inference, prompt chaining, and data processing pipelines. The backend exposes clean REST APIs consumed by the Flutter app, keeping the mobile layer lightweight.

Python (FastAPI)OpenAI / LLM APIsCloud-hosted GPU inference

Authentication & Data Layer

Firebase Authentication handles secure user sessions with role-based access. User data, interaction logs, and AI response history are stored in Firestore with strict security rules, ensuring privacy and fast reads.

Firebase AuthCloud FirestoreFirebase Security Rules

Performance & Latency

Sub-500ms AI Response

The app is optimized for speed — from request to AI response in under 500ms. This is achieved through request batching, response caching for repeated queries, and co-located inference servers reducing network round-trip time.

Response cachingRequest batchingOptimized model quantization

Real-Time Architecture

End-to-End Data Flow

Flutter App → Firebase Auth → REST API (Python Backend) → AI Inference Engine → Firestore (result storage) → FCM (push updates). The entire pipeline is async and non-blocking, keeping the UI responsive at all times.

FlutterFirebase AuthFastAPIPython AI ModelsFirestoreFCM

Technologies

Flutter

Dart

Python

iOS

Firebase

FastAPI

OpenAI API

Firestore

Project Screenshots