Senior Staff Software Engineer (Product ML Infrastructure)

Join Pinterest as a Senior Staff Software Engineer in the Product ML Infrastructure team. In this role, you will set the technical vision and roadmap for model training and serving, lead architectures for distributed training and high-scale inference, and improve efficiency across various aspects of the infrastructure. You will also partner with Ads and Core AI/ML teams to productionize features and models, drive cross-organizational architecture decisions, and mentor senior engineers. The ideal candidate will have deep expertise in distributed ML systems, strong GPU performance knowledge, and experience with AI/ML modeling and systems programming.

• Set the technical vision and roadmap for model training and serving across PMLI, with reusable interfaces to data and feature infrastructure.
• Lead architectures for distributed training, fine-tuning, distillation, evaluation, and high-scale CPU/GPU inference.
• Drive cross-organizational architecture decisions, migrations, and operational standards; mentor senior engineers and raise the engineering bar.

• Deep expertise in distributed ML systems, including production experience with both large-scale training and online inference
• Bachelor’s degree in Computer Science, Engineering, a related field, or equivalent experience
• Experience with AI/ML modeling, recommender systems, Ads ranking, retrieval, feature platforms, or similarly demanding ML workloads
• Strong GPU performance knowledge, such as profiling, distributed execution, kernel and memory optimization, compilation, or quantization
• Strong systems programming and design skills in C++, Java, or Python
• High ownership and sound judgment in reliability, security, cost, and operational excellence
• Demonstrated ability to use AI to improve speed and critically evaluate AI-assisted work, with accountability for correctness, quality, and sensitive data
• A track record of setting technical strategy and delivering company-wide infrastructure initiatives in ambiguous environments