
Staff ML Research Engineer, Multimodal Structure & Marengo
TwelveLabs9 hours ago
Seoul, Korea, SouthStaff+
Responsibilities
- Set technical direction for multimodal structure and define how video, audio, text, and document assets become reusable addressable units.
- Define architectures, training strategies, and data strategies for next-generation multimodal embedding models.
- Own model development from research planning through distributed training, production evaluation, and iteration.
- Architect and optimize distributed training pipelines, data-processing systems, experiment workflows, and GPU utilization.
- Own production model APIs, including packaging, API design, inference optimization, and reliable operation at scale.
- Build large-scale data curation, filtering, and quality systems for structure and embeddings.
- Define evaluation methods and quality standards for structure, embeddings, and end-to-end multimodal understanding and retrieval.
- Drive the interface between multimodal structure and semantics and align model integration with Agent, Search, Product, and Infrastructure teams.
- Raise the research engineering bar through design reviews, experiment reviews, and technical mentorship.
Requirements
- 7+ years of industry experience in computer vision, video understanding, or multimodal learning.
- Demonstrated ability to turn ambiguous research problems into concrete, impactful solutions from problem identification through delivery.
- Deep expertise in large-scale distributed model training, including kernel optimization, FSDP, or similar approaches.
- Experience building and operating production model APIs for large-scale machine-learning systems.
- Deep expertise in video understanding, multimodal representation learning, or foundation-model development.
- Experience building end-to-end systems connecting multimodal structure with embeddings and retrieval.
- Proven ownership of defining, building, deploying, and iterating on ML systems in production.
- Strong proficiency in Python and PyTorch.
- Evidence of both research depth and engineering impact, such as publications paired with shipped products.
- Experience training models at billion-parameter scale, operating reliable training pipelines, and building large-scale data curation and quality systems is preferred.
- Preferred experience includes temporal, spatial, or hierarchical modeling; multimodal video modeling; training and inference optimization; and technical leadership over architectural decisions.
- An MS and deep industry experience are described as a typical strong fit, but the posting does not state them as an absolute requirement.
Benefits
- Hybrid work with autonomy and collaboration.
- MacBook and approximately KRW 700,000 in work-from-home equipment support, with equipment replacement every three years.
- Unlimited LLM tokens for technology employees using AI responsibly and efficiently.
- Annual professional-development allowance of approximately KRW 1,400,000 for courses, conferences, and memberships.
- English education and global buddy programs.
- Taxi reimbursement for late-night and weekend commuting.
- Annual corporate card allowance of approximately KRW 7,200,000 usable for meals, transportation, and other expenses.
- Office snack bar with snacks, coffee, and seasonal fruit, plus dinner support after 7 p.m. when working in the office.
- Annual health screening for the employee and one family member, group insurance, and influenza vaccination support.
- Two-week paid holiday break at year-end.
- Three-month probation period with full salary; the hiring process may include in-person interviews.
Categories
AI ResearchML Engineering
About TwelveLabs
TwelveLabs builds video-understanding infrastructure and multimodal AI models that let developers and enterprises search, analyze, and build applications from video and audio. It sells APIs and platform integrations, working with major clouds (AWS, Azure, GCP) and data platforms (Databricks, Snowflake) to support production-scale workloads in media, entertainment, sports, security, and government. Founded in 2021 and headquartered in San Francisco, it is privately held.