We are looking for a Research Scientist to join the Creative Vision Research Team! At Creative Vision, we focus on making everyone into a creator. We believe creativity is achieved when technology understands the world, humans and objects, provides a range of creative generation and manipulation tools and efficient real-time experiences. Our technology impacts and contributes to multiple products at Snap. 

What you’ll do:

• Define, lead and execute a multi-year research agenda

• Architect and implement large-scale multimodal generative projects

• Share your expertise with other teammates and interns

• Publish your work to top conferences

• Partner with engineering teams to deliver your technology to millions of Snapchatters

Knowledge, Skills, & Abilities:

• Strong technical knowledge of statistics, machine learning, vision and state-of-the-art deep learning literature

• Proven record of defining, leading and executing challenging research projects

• Strong computer science fundamentals, problem solving skills, programming (Python, C, C++)

• Proven ability to lead interns, PhD students, junior researchers or ML engineers

Minimum Qualifications: 

• PhD in a related technical field such as computer science, statistics, mathematics, machine learning or equivalent years of experience 

• 2+ years of industry or postdoc experience with large scale generative AI, building foundational models, and large model training in academia or industry

• Experience with large scale, distributed training across hundreds of machines and cloud computing

• Experience with large-scale data collection and annotation best practices

• Track record of publications in top-tier international research venues (e.g. ICLR, AAAI, NeurIPS, CVPR, ECCV, ICCV, SIGGRAPH) 

Preferred Qualifications: 

• Experience with multimodal machine learning, involving images, videos, audio, text, speech

• Experience with 3D computer vision, non-rigid and articulated 3D registration and synthesis

• Experience with efficient Gen AI techniques, such as quantization, pruning, distillation, efficient training for generation and understanding, parameter- and data-efficient learning, and efficient VLMs

• Experience with low level vision techniques, such as image and video quality assessment, compression, restoration, and super-resolution

• Experience with 3D Articulated human motion generation 

• Experience with CUDA framework and low level optimization of generative pipelines for training and inference

• Strong theoretical background in generative modeling and diffusion models