Top Tech Jobs & Startup Jobs in San Francisco Bay Area, CA

Design and implement algorithms, model architectures, data mixtures, and scaling laws for large-scale foundation model training. Run and oversee large distributed training experiments, collaborate with infrastructure, data, and multimodality teams, and apply rigorous, data-driven ablations to drive flagship training runs and system improvements.

Build and optimize inference systems and tools for generative AI models, collaborate with researchers to improve performance, profiling, debugging, and deploy large-scale distributed inference solutions.

Design and build petabyte-scale multimodal data pipelines and infrastructure (text, audio, images, video). Own and maintain systems like Spark, Ray, and vector databases, partner with pre/post-training teams to improve data recipes through experimentation, and support large-scale dataset storage and processing for AI model training.

Design, implement, and optimize distributed pre-training infrastructure for large-scale GPU clusters. Profile, benchmark, and debug compute, memory, networking, and storage bottlenecks; optimize communication libraries and GPU programming stacks; collaborate with hardware and model teams to scale training, ensure reliability, and influence architecture and roadmap.

Contribute to building and scaling ML development infrastructure for Copilot: design and optimize CI/CD for large-scale training, build developer tools, maintain multi-cloud infrastructure, and deploy/host models for inference and data generation while collaborating with cross-functional teams to improve reliability, performance, and model development workflows.

Design and build compute orchestration and scheduling on Kubernetes and Ray for large-scale AI model training and inference. Tune and benchmark pretraining software on Nvidia GB200 and AMD MIxxx GPU clusters, improve cluster efficiency, enable observability and reliability, collaborate with research and framework teams, and influence architecture and roadmap for Copilot and foundational models.

Develop and iterate post-training methods for large language models: drive data collection, build evaluations, design experiments, and apply reward modeling and RL techniques to improve model capabilities and deployment.

Design, curate, and evaluate high-quality multimodal datasets for foundation models. Build scalable ingestion, preprocessing, filtering, annotation pipelines; run data ablations and analyses; create auditing, visualization, and versioning tools; collaborate with Safety, Ethics, and research teams to ensure data quality, privacy, and responsible AI alignment.

Design and develop multimodal foundation-model algorithms and architectures, run large-scale experiments, improve training and deployment efficiency, build datasets and pipelines, perform rigorous evaluation and ablation studies, and collaborate with infrastructure, data, and product teams to responsibly advance AI systems.

Drive product strategy and KPIs for backend networking of ultra-large AI supercomputers. Identify failure points, lead remediations and roadmap changes, and coordinate cross-functional teams including customers, datacenter operators, vendors, and product groups to maximize uptime and AI workload throughput.