Pinterest is looking to develop and improve its internal text-to-image generative model, Pinterest Canvas, to enhance visualization, inpainting, and outpainting products. This involves leveraging rich visual-text datasets and developing foundation ML models to improve the core product.
Requirements
- Hands-on experience working with diffusion text-to-image models and independent model implementation skills.
- Experience working with generative computer vision models, preferably various forms of diffusion models.
- 5+ years of industry computer vision experience.
- Publications at top ML conferences.
Responsibilities
- Prototype new model architectures for Pinterest Canvas, our internal text-to-image generative model.
- Read research papers, participate in group discussions, and help brainstorm our overall visual generative strategy at the company.
- Help with collection of relevant visual training data for Pinterest Canvas, particularly to conduct RLHF, targeted fine-tuning, etc.
- Publish and publicize your work via conferences, paper submissions, blog posts, etc.
- Mentor more junior researchers or research interns within the Pinterest Labs organization.
- Develop foundation ML models that fully leverage the tens of billions of Pins and the associated knowledge graph to improve the core product.
- Participate in the development of our core multimodal text & image embeddings, as these are in turn used to condition the model in unique ways.
Other
- M.S. or PhD in Machine Learning, Computer Science, or related areas.
- This role will need to be in the office for in-person collaboration 1-2 times/quarter and therefore can be situated anywhere in the country.
- This position is not eligible for relocation assistance.
- US based applicants only
- The position is also eligible for equity.
