Staff Machine Learning Engineer, Visual AI

About the Team:

Hundreds of millions of people around the world come to our platform to find creative ideas, dream about new possibilities and plan for memories that will last a lifetime. At Pinterest, we’re on a mission to bring everyone the inspiration to create a life they love.

Within Pinterest, the Pinterest Labs organization focuses on applied ML research and development to power the platform. Labs works across a broad variety of AI/ML initiatives, including LLMs/VLM, agent design, core computer vision, multimodal representation learning, visual generative modeling, recommender systems, graph learning, and more. This is the group that develops the foundation AI models that fully leverage the hundreds of billions of Pins and the associated knowledge graphs, and ships new product capabilities to fully utilize these technologies.

We are currently hiring for the Visual team in Labs, which develops Pinterest's foundation visual models. In this role, you'll work with Pinterest's rich visual-text dataset to train large-scale VLMs, encoders, and diffusion models from scratch that are continuously shipped to production to power visualization and search capabilities. The team is subdivided into two pods, which share pretraining datasets, training and RL infrastructure, and generally co-develop our visual modeling ecosystem. The visual understanding pod builds the core visual embeddings and token models such as PinCLIP and deeply integrates them with the VLM/LLM/agent ecosystem at Pinterest. The visual generative pod builds Pinterest Canvas, our production image editing and generation model used across our assistant, visualization, and monetization products. Labs is staffed to be a highly collaborative environment where research scientists, ML engineers, product builders, and infrastructure engineers all work directly together, so that the team can plug into any technical effort at the company to support fast productionization.

What you’ll do:

• Build state-of-the-art visual encoders, VLMs, and diffusion models that power Pinterest's visual AI capabilities
• Experiment with billion-scale image datasets, backed by large-scale GPU computing.
• Build flexible visual reasoning tools such as composed image retrieval, promptable image feature computation, instruction-tuned embedding and generative editing models, and more.
• Read research papers, participate in group discussions, and directly participate in brainstorming the company's overall AI strategy.
• Help construct data agents to build training data that can be shared across multimodal representation, composed image retrieval, image-editing generation, and visual language modeling.
• Collaborate directly with product engineers and infrastructure engineers to ship new capabilities in the core product.
• Publish and share your work through conferences like CVPR and KDD, paper submissions, and blog posts.
• Mentor junior researchers and research interns within the Pinterest Labs organization.
• Collaborate across a team situated across San Francisco, Seattle, NYC, and remote roles.

What we’re looking for:

• Research engineers and scientists with experience building and training large scale vision models of all categories.
• Experience with multimodal representations and visual language modeling is strongly preferred.
• A track record of research contributions (e.g., publications, open-source work) and/or shipping ML models to production.
• Hands-on experience with large-scale model training and modern deep learning frameworks (e.g., PyTorch).
• Strong collaboration skills and a demonstrated ability to work effectively in a small, fast-moving team.
• M.S. or PhD in Machine Learning or related academic areas, or equivalent work experience.
• Experience using AI-accelerated research tooling akin to auto-research, data agents, etc.

Relocation Statement:

• This position is not eligible for relocation assistance. Visit our PinFlex page to learn more about our working model.

In-Office Requirement Statement:

• We let the type of work you do guide the collaboration style. That means we're not always working in an office, but we continue to gather for key moments of collaboration and connection.
• 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.

AI use statement:

• Through our interview process, what matters most is that you can always explain your approach, showing us not just what you know, but how you think. You can read more about our AI interview philosophy and how we use AI in our recruiting process here.