Design, implement, and scale state-of-the-art LLM's core image/video understanding capabilities. Drive end-to-end experimental workflows across pre-training and post-training. Build robust, scalable infra and data pipelines to support high-quality multimodal training and eval benchmarks. Partner closely with Research Scientists and SWEs to transition cutting-edge research prototypes into production-ready foundation models, striking the right balance between rapid prototyping, maintainability, and generality. Analyze complex experiment metrics and failure modes across multimodal benchmarks to identify opportunities for model quality hillclimbing.
Job description provided by Google.
View the original posting
