Yutori
AI Engineer — LLM Infra
1y ago
USASeniorml infrastructuregpu clustersncclinferencepost-trainingrl
Build and scale infrastructure for multimodal large language models and superhuman AI agents on the web.
Responsibilities
- Scale infra for post-training of multimodal LLMs (CPT, SFT, RL, search, reward models)
- Scale infra for agentic inference (throughput and latency of perception-planning-action loops)
- Build the foundations of a superhuman generalist web-agent
- Work closely with product engineers to translate cutting-edge AI capabilities into reliable product experiences.
Conditions
- Competitive salary and equity
- Visa sponsorship and relocation stipend to bring you to SF
- Generous health, dental, vision insurance for you and your dependents
- 20 days of paid time off per year
- Work laptop and budget to set up your work office
- Daily team lunches
- Commuter benefits
- Small, focused team of high-potential individuals. In-person in SF.
Other
- Yutori is reimagining how people interact with the web by building AI agents that can reliably do everyday digital tasks. We are building the entire stack to be agent-first, from training our own models to generative product interfaces.
- Towards this goal, we are looking for a member of the AI technical staff to join the founding team. Someone technically strong, and excited about building superhuman AI agents that take actions on the web.
- Our founders — Devi Parikh, Abhishek Das, Dhruv Batra — have decades of experience in AI research and product spanning generative, multimodal and embodied AI at Meta. Our team combines AI experience with design-minded product thinking to build and deliver on Yutori’s mission.
- Yutori is backed by a stellar set of visionary investors — Elad Gil, Sarah Guo, Jeff Dean, Fei-Fei Li, Amjad Masad, Guillermo Rauch, Akshay Kothari, Soleio, Oliver Cameron, Julien Chaumond, Logan Kilpatrick, Bryan McCann, Vladlen Koltun, Jamie Cuffe, Michele Catasta, etc.
- Experience with ML infrastructure (GPU clusters) and supporting networking (NCCL)
- Experience optimizing post-training and inference performance of multimodal LLMs (data/tensor/pipeline/context/expert parallelism, optimizing MFU, throughput, latency)
- Low level systems experience (Triton, CUDA)
- High IQ, high EQ, high agency, high craftsmanship, low ego. Proactive, clear communication.