Research Scientist, Video & Multimodal

Location: Remote - United States

Salary: $160,000–$185,000 per year, based on experience, skills, and qualifications.

Company

hirify.global is a global data engineering company that provides data, evaluation frameworks, platforms, and human expertise for generative AI builders and adopters.

What you will do

• Define data specifications, annotation schemas, sampling strategies, and evaluation criteria for video and multimodal models.
• Develop evaluation methodologies for video understanding, including temporal grounding, long-context reasoning, event localization, retrieval, and cross-modal evaluation.
• Develop evaluation methodologies for video generation focused on fidelity, temporal coherence, physical plausibility, and human judgment.
• Run fine-tuning, evaluation, and ablation experiments to measure the impact of data decisions.
• Design adversarial and stumping evaluations, convert model failures into improved data, and publish benchmarks, methodologies, and research papers.
• Collaborate with annotation teams, subject-matter experts, customers, frontier labs, and synthetic-data specialists to operationalize collection and labeling plans.

Requirements

• Approximately 5+ years of hands-on industry experience in video understanding or multimodal machine learning; a current PhD research agenda may offset experience at the lower end.
• A Bachelor's degree in computer science, electrical engineering, or a related technical or quantitative field is required; an MS or PhD is preferred.
• Hands-on experience training and evaluating video or multimodal models, with strong PyTorch fundamentals.
• Experience with ffmpeg, decord, temporal and COCO-style annotations, WebDataset, Parquet, Arrow, and HuggingFace datasets.
• Experience fine-tuning large video or vision-language models using HuggingFace Transformers, PEFT, and efficient inference, including work with long-form video, streaming, temporal segmentation, or synthetic video generation.
• First-author publications or strong open-source contributions at venues such as CVPR, ICCV, ECCV, NeurIPS, or ICLR, plus the ability to explain technical decisions clearly and conduct reproducible experiments.

Nice to have

• Interest or hands-on experience in responsible-AI evaluation, red-teaming, safety testing, or robustness testing for video and multimodal systems.

Culture & Benefits

• Direct collaboration with customers and frontier labs developing video understanding, video-language, and video-generation models.
• Practical research focused on data quality, evaluation science, benchmarks, and measurable model improvements.
• Opportunity to publish research that advances video and multimodal AI.
• Remote work from the United States.