Pinterest

The Visual Search team works on all aspects of applied computer vision at Pinterest and is offering summer 2021 internship opportunities, along with contracting roles for more experienced developers who are looking for a flexible arrangement. This is a full stack team, ultimately driven to ship user-facing products, like Pinterest Lens, Visual Shopping, and Pinterest AR. We do this by building state of the art computer vision signals, such as large scale multi-task embeddings, high-throughput detection models, generative adversarial models for image composition, and many others. These models and computer vision features power product experiences on Pinterest for hundreds of millions of users. Pinterest Visual Search team runs on PyTorch, C++, and lots of Python, among with many other technologies. We encourage publication of all of our work (CVPR, KDD, etc.), and have a highly collaborative and flexible work environment. We’re looking for roles right now: interns with an interest in shipping computer vision products, and contractor roles for more experienced engineers to help us build better ML data tooling and visualizations. Interns: If you’re passionate about product or computer vision infrastructure, such as building label-to-deployed-model toolchains via active learning, or if you’ve had experience with interesting CV challenges, or you are just looking for a highly applied ML work, reach out to our team! You can contact me directly by pinging dkislyuk@pinterest.com with the tag [VISUAL SEARCH HN] in the title. Please highlight a CV project or an interesting product you've worked on to help our team figure out what role to consider you for. Any ML infra or tools that you are interested in are a big plus. Contractors: If you have experience with data workflows for ML projects, or building dataset visualization front-ends, or have some other expertise that you think is relevant to the problems above, send me a message (same deal, please add the tag [VISUAL SEARCH HN]). The position would be fully remote.