Research Scientist, Video & Multimodal

The role

You will own the evaluation science for video understanding and video generation models. You will define how video data is designed, structured, and evaluated to support multimodal models.

What you will do

• Translate video and multimodal model requirements into data specifications
• Build evaluation methodology for video understanding
• Build evaluation methodology for video generation
• Decide how video should be structured, enriched, and sampled
• Run experiments to prove data decisions matter
• Design adversarial and stumping evaluations
• Publish benchmarks, methodology, and papers
• Work with annotation teams and subject-matter experts

Required qualifications

• Roughly 5+ years of hands-on industry experience in video understanding or multimodal ML
• A PhD with a compelling, current research agenda can offset the lower end of experience
• Bachelor's degree in computer science, electrical engineering, or a related technical or quantitative field
• Trained and evaluated video or multimodal models yourself
• Strong PyTorch fundamentals
• Fluency in ffmpeg and decord pipelines, temporal and COCO-style annotation, WebDataset, Parquet and Arrow, and HuggingFace datasets
• Experience fine-tuning large video or vision-language models with the modern toolchain (HuggingFace transformers, PEFT, efficient inference)
• Experience with long-form video, streaming, temporal segmentation, or synthetic video generation
• Experience building evaluation sets and calibrating difficulty
• First-author publications or strong open-source contributions at venues such as CVPR, ICCV, ECCV, NeurIPS, or ICLR
• Ability to work directly with research scientists at customers and frontier labs
• Ability to explain data and modeling decisions clearly to expert and non-expert audiences

Nice to have

• Advanced degree (MS or PhD) in a relevant field
• Interest or hands-on experience in responsible-AI evaluation and red-teaming

About the company

Innodata is a global data engineering company. Their mission is to enable the responsible advancement of artificial intelligence by providing data, evaluation frameworks, and human expertise.