• Brought proof-of-concept clothing recognition system to production quality and scalability

• Developed retrieval indexes for several clothing brands (to find the most similar garments to a query image)

• Improved neural network design and usage, greatly reducing false positive matches (using DNNs with Caffe)

• Assisted with web-crawling, writing scripts to gather and curate labelled datasets

• Optimized code for processing queries and design of retrieval index to handle queries in ~100ms