AWS Cloud Album Project

I designed and implemented a cloud-native photo management system for scalable storage, asynchronous processing, automated classification, and metadata-based search. It can manage large photo collections and display images by location.

Core features

• Image classification: YOLOv8 runs through ONNX and stores labels with confidence scores for search and organization.
• Asynchronous processing: SQS and Lambda separate uploads from ML inference, keeping user requests responsive and allowing workers to scale independently.
• Storage: image files live in Amazon S3; users, labels, locations, processing state, and other metadata are stored in SQL.
• Backend: Flask REST APIs handle authentication, upload, retrieval, and album management. JWT and IAM-based controls protect access.
• Map view: EXIF GPS data is extracted and shown with OpenStreetMap and Leaflet, with location-based browsing and filtering.
• Follow-up metadata: the system can ask for missing information after upload to make later searches more accurate.

Architecture

• Frontend: HTML and CSS client
• Backend: Flask API on EC2
• Async layer: SQS queue and Lambda workers
• Storage: S3 for images and SQL for metadata
• ML: YOLOv8 inference through ONNX
• Maps: OpenStreetMap rendered with Leaflet

The main engineering work was connecting storage, queues, serverless workers, inference, and relational metadata into one event-driven pipeline, while retaining explicit processing status and secure communication between services.