Job Purpose:
MBZUAI is looking to recruit a Research Engineer to design and
maintain scalable video data pipelines, develop robust
preprocessing and annotation workflows for world model training,
and optimize distributed systems supporting multimodal foundation
models. The successful candidate will enhance inference efficiency
for real-time interaction, integrate research prototypes into
production-ready systems, and contribute to rigorous quantitative
evaluations of model performance. Strong coding skills, experience
with large-scale infrastructure, and close collaboration with
research teams to translate experimental ideas into reliable
engineering solutions are essential.
Key Responsibilities:
- Design, implement, and maintain scalable video data pipelines
to support large-scale training.
Develop data preprocessing, transformation, and synthesis
workflows to support world model training.
Contribute to building high-quality data annotation pipelines
to ensure accurate and consistent labels across large-scale
datasets.
Support the training of multimodal foundation models (e.g.,
video diffusion models, world models) by developing and optimizing
distributed training systems.
Improve inference and serving efficiency for real-time
interaction through model optimization and system tuning.
Monitor system health and performance and contribute to
debugging and optimization at scale.
Work closely with research teams to understand experimental
goals and translate ideas into reliable and maintainable
infrastructure and tools.
Integrate novel research prototypes into production-ready
systems and ensure reproducibility at scale.
Participate in design and code reviews, ensuring code quality,
efficiency, and compliance with best practices.
Contribute to the development of tools and infrastructure to
evaluate model performance using rigorous quantitative benchmarks,
including metrics for physical accuracy and controllability.
Maintain and extend shared codebases, contribute to internal
documentation, and support onboarding of new team members or
collaborators.
Write clean, efficient, and well-tested code for components
across the model development lifecycle.
Support contributions to research papers and demos when
engineering work plays a significant role.
Help represent the team’s engineering excellence in internal
and external forums when appropriate.
Academic Qualifications:
- MSc or PhD in Machine Learning or Computer Science, or
equivalent industry experience.
Professional Experience:
- Proficient in data collection, cleaning, and transformation at
scale, including designing robust pipelines for multimodal datasets
(e.g., video, audio, text).
- Practical experience with web scraping and crawling frameworks
(e.g., scrapy, selenium, playwright, BeautifulSoup) to collect and
curate high-quality web-scale datasets.
- Experience in large-scale model training (LLMs or Diffusion
Models) on large clusters.
- Hands-on experience with state-of-the-art video generative
models (e.g., Sora, Veo2, MovieGen, CogVideoX, etc.).
- Experiences in building and optimizing large-scale video data
pipelines.
- Experience in accelerating diffusion model inference for
improved efficiency.
- Exceptional problem-solving and troubleshooting skills to
tackle complex technical challenges.
- Strong systems and engineering expertise in deep learning
frameworks such as PyTorch.
- Strong communication and collaboration skills for effective
cross-functional teamwork.
- Demonstrated ability to solve complex system-level challenges
and debug failures across the training/inference stack (e.g.,
memory issues, deadlocks, I/O bottlenecks).
