SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Twelve Labs is building the intelligence layer for video understanding at scale. The company's multimodal AI models understand video across sight, sound, and motion, powering production workloads for media, entertainment, sports, security, and government. Twelve Labs has raised over $210M from top-tier investors including NEA, Amazon, NVIDIA, Snowflake, and Databricks, with offices in San Francisco, Seoul, New York, and London.
The Multimodal Structure & Embeddings team develops core capabilities for understanding video, audio, text, and documents. The team owns the entire model development lifecycle: building large-scale training datasets, designing architectures, optimizing distributed training, and developing evaluation frameworks. Marengo, Twelve Labs' multimodal embedding model, represents units of content in a shared embedding space for retrieval and understanding.
As a Staff ML Research Engineer, you will set technical direction for next-generation multimodal models and own end-to-end development from research strategy through production deployment. This is a high-autonomy role at the intersection of video understanding, multimodal representation learning, large-scale systems design, and cross-team technical leadership.
Key responsibilities include: defining technical direction for multimodal structure and how assets are organized into reusable units; architecting training and data strategy for next-generation embedding models; owning end-to-end model development from research planning through distributed training to production evaluation; designing and optimizing large-scale training infrastructure including distributed pipelines, data processing, and GPU utilization; owning production model APIs from packaging and design through inference optimization and reliable operation at scale; driving data strategy through curation and quality systems; and defining evaluation methods and quality standards.
You will work with world-class compute infrastructure including NVIDIA B300 GPUs, enabling rapid iteration on large-scale experiments. The path from research to production is exceptionally short, with close collaboration across Agent, Search, Product, and Infrastructure teams to improve models serving thousands of customers worldwide.
