Image Recommendation Platform
A critical goal of content sites is to keep users engaged for an extended period of time. With this in mind, we have extracted GIF view data from user sessions over the past two years and built a recommendation system that suggests similar GIFs based on users’ areas of interest.
This system is able to “understand” the meaning within selected graphic imagery by examining users’ intent and behavior. The platform makes tailored content suggestions and provides a wider selection of GIFs that are likely to appeal to the particular user. We have also implemented a technology for continuous training which updates the system so that it can process new training data automatically.
- Python
- Pytorch
- Google BigQuery
- Fast Approximate Nearest Neighbor Search
- AWS
- Sanic
- Numpy
- Nmslib
- Metric Learning
- Kubernetes
- Docker
- Word2Vec
- Sparse Vector Spaces
- Continuous Learning
- Jenkins