Video & Multimodal AI Research Scientist

Innodata; Ridgefield Park (NJ); On-site; USD 160,000 - 185,000; Full time

Job summary

Innodata is hiring a Research Scientist to own the science behind video and multimodal evaluations. You will define data designs, evaluation criteria, and experimental protocols to advance video understanding and generation.

The role emphasizes rigorous experiments, cross‑functional collaboration, and publishing benchmarks and methodologies. Required are 5+ years in video or multimodal ML, a CS/EE degree (MS/PhD preferred), and strong PyTorch experience.

Qualifications

• 5+ years of hands-on industry experience in video understanding or multimodal ML.
• Bachelor's in CS/EE or related technical field required; MS/PhD preferred.
• Proficient with PyTorch and model fine-tuning in video/vision-language tasks.

Responsibilities

• Define data design, schemas, and evaluation criteria for video and multimodal models.
• Create evaluation methodologies for temporal grounding, long-horizon reasoning, and cross-modal retrieval.
• Develop data collection and labeling plans with annotation teams and synthetic-data pipelines.
• Run experiments linking data choices to model performance and publish results.

Skills

Video understanding

Multimodal ML

PyTorch fundamentals

HuggingFace transformers

Datasets & benchmarks

Education

Bachelor's in CS/EE or related

MS or PhD preferred

Tools

ffmpeg

decord

WebDataset

Parquet/Arrow

HuggingFace datasets

Job description

Innodata is hiring a Research Scientist to own the science behind video and multimodal evaluations. You will define data designs, evaluation criteria, and experimental protocols to advance video understanding and generation.

The role emphasizes rigorous experiments, cross‑functional collaboration, and publishing benchmarks and methodologies. Required are 5+ years in video or multimodal ML, a CS/EE degree (MS/PhD preferred), and strong PyTorch experience.