Staff ML Research Engineer, Multimodal Structure & Marengo
Core
Set technical direction and own end-to-end development for next-generation multimodal embedding models and structure systems that organize video, audio, and text for retrieval and understanding.
Role type
Staff ML Research Engineer (Multimodal Structure & Embeddings)
Builds
Production-grade multimodal AI models (Marengo, Pegasus) powering Jockey, an agentic system for corpus-level video understanding and retrieval.
Domain
AI/ML, Computer Vision, Video Understanding, Multimodal Learning
Deliverable
production ML models
Required skills
Computer vision, video understanding, multimodal representation learning, large-scale distributed model training, production model APIs, Python, PyTorch
Preferred skills
None explicitly stated
Technologies
NVIDIA B300 GPUs, PyTorch, FSDP
Responsibilities
Set technical direction for multimodal structure; define architecture and data strategy for embedding models; own end-to-end model development lifecycle; architect and optimize large-scale training infrastructure; own production model APIs; drive data strategy for curation and quality; define evaluation methods; drive cross-functional alignment; raise research engineering bar through mentorship.
Seniority
Staff, high-autonomy IC with cross-team leadership