Research Scientist, Video & Multimodal

What you'll do

• Design video-data specifications and evaluation methods, including temporal grounding, long-form reasoning, generation quality, and human versus model-based scoring.
• Run fine-tuning and evaluation experiments, build adversarial evaluations, and use findings to improve data and benchmarks.
• Publish research and coordinate with customers, frontier labs, annotation teams, subject-matter experts, and synthetic-data teams.

What they ask for

• Bachelor's degree in computer science, electrical engineering, or a related technical or quantitative field.
• Hands-on training and evaluation of video or multimodal models, with strong PyTorch fundamentals.
• Experience with video-data formats and tooling, including ffmpeg, decord, WebDataset, Parquet and Arrow, and HuggingFace datasets.
• Experience fine-tuning large video or vision-language models with modern tooling, and a track record of first-author publications or strong open-source contributions.

Nice to have

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

Pay

$160,000 - $185,000 USD base salary

Tools

PyTorch, ffmpeg, decord, WebDataset, Parquet, Arrow, HuggingFace datasets, HuggingFace transformers

AI focus

Research and evaluation for video understanding, video generation, and multimodal models, including temporal reasoning, grounding, cross-modal evaluation, and responsible-AI testing.