Technical Case Study & System Architecture
Users often see products in the real world but struggle to find the exact or similar items on local e-commerce platforms using text search.
AI/Fullstack Engineer. Developed the computer vision pipeline and the web scraping infrastructure to build the product catalog.
Leveraged TensorFlow to generate product embeddings, stored in a vector-capable database to enable sub-second visual similarity matching.
Built a resilient scraping system using Python to aggregate product data from multiple local vendors, handling rate limits and dynamic content.
Microservice-oriented with a Python-based AI core and a Next.js frontend, communicating via RESTful APIs.
PostgreSQL for cataloging scraped metadata and TensorFlow-processed vector embeddings for visual search.
Standard JWT authentication for user profile persistence and personalized recommendations.
Distributed scraping nodes and GPU-accelerated inference workers for real-time image processing.
"Created a functional proof-of-concept that demonstrates the power of AI in bridging the gap between physical and digital retail."
Gained significant experience in scaling AI inference and managing large-scale data ingestion pipelines.