Back to ProjectsDental Caries Detection with AISee the full case study: How Claude powers DexaVision

    DexaVision

    AI dental caries detection platform that analyzes images using ICDAS II classification, connects patients with verified dentists via real-time chat, and processes subscription payments. All from a mobile device.

    < 30sFull AI analysis
    4 modelsMulti-model pipeline
    45+Lambda functions
    DexaVision

    Overview

    How the system works

    DexaVision is a complete dental health platform combining a multi-model AI pipeline with real-time consultations. Patients upload dental photos from the mobile app and an AI pipeline in AWS Step Functions executes: image validation, similar case search with RAG, ICDAS 0-6 classification by specialized AI models, and final validation by an AI arbiter. Doctors verify with documents, manage availability, and handle consultations via real-time chat. Admins supervise verifications and analyses from a web dashboard. In-App Purchase (IAP) subscriptions and push notifications.

    Mobile AppAWSStep FunctionsBedrockMulti-Model AIDynamoDBReal-time ChatIAP
    The Challenge

    The Challenge

    Caries diagnosis requires an in-person visit, long wait times, and manual evaluation. Patients in rural areas lack easy access and dentists spend hours on initial reviews that could be automated.

    The Solution

    The Solution

    Multi-model AI pipeline in AWS Step Functions: image validation → RAG with curated knowledge base (36 cases) → specialized AI caries classifier → validating AI arbiter. Real-time chat, doctor verification system, IAP subscriptions, and push notifications.

    Results

    Results

    Complete analysis in under 30 seconds with ICDAS II classification validated by multiple models. Real-time doctor-patient consultations with auto-acceptance. 45+ Lambda functions, 5 DynamoDB tables with KMS encryption, HIPAA compliance with 7-year retention.

    Screenshots

    Discover the project

    Technical Architecture

    Technology Stack

    Native mobile app (iOS/Android)
    Web admin dashboard
    AWS infrastructure (fully automated)
    Step Functions for AI analysis pipeline
    AWS Bedrock with specialized AI models
    REST API + real-time chat
    DynamoDB (5 tables) with encryption
    S3 + CloudFront for dental images (HIPAA 7-year retention)
    Authentication and push notifications
    IAP (In-App Purchase) for subscriptions
    Async processing with retry handling

    Key Benefits

    Improvements after implementation

    Full AI analysis

    Full AI analysis

    < 30s — Full AI analysis

    Multi-model pipeline

    Multi-model pipeline

    4 models — Multi-model pipeline

    Lambda functions

    Lambda functions

    95% reduction in initial diagnosis time. 24/7 access without travel for rural or remote patients.

    Quote Icon
    The combination of multiple AI models specialized at each diagnosis stage gives us precision that we wouldn't achieve with a single model. Patients receive results in seconds and can consult with a verified dentist without leaving the app.
    DexaVision TeamDevelopment, DexaVision
    Case Studies

    Discover more case studies

    Sugeter

    Plan, share and discover travel itineraries in minutes

    Social Travel App

    Instagram-style social platform for travelers where users can create itineraries, add activities, locate points on a map, and share experiences in a guided, fluid, and frictionless flow.

    70%

    Time reduction

    100%

    i18n Ready

    0→1

    Scalable architecture

    View landing pageexternal

    Lancémonos Nomás

    276+ videos processed with AI, searchable in milliseconds

    AI-Powered Educational Platform with HLS Streaming

    Video course platform with automatic AI transcription, semantic topic analysis, intelligent search (Algolia), and HLS streaming that solves memory issues with long videos. 276+ videos processed in production.

    276+

    Videos processed

    < 200ms

    Algolia search

    -90%

    Indexing time

    View full caseexternal