Lead Research Scientist, Multimodal Foundation Models

Pika; Palo Alto (CA); Hybrid; USD 190,000 - 250,000; Full time

Benefits offered by this job

Competitive salary

Equity

Health benefits

Hybrid work option

Job summary

Pika is seeking a Research Scientist to lead the development of large-scale multimodal foundation models, focusing on pre-training and mid-training. You will design novel algorithms, curate data, and drive production-ready implementations in collaboration with engineering and product teams.

Ideal candidates have 5+ years of research experience, strong publication records, and hands-on expertise with PyTorch/TensorFlow.

Qualifications

• 5+ years of research experience in large-scale pre-training/mid-training of multimodal foundation models.
• First-author publications in top conferences or journals (e.g., NeurIPS, ICML, ICLR).
• Hands-on experience with design, training, and deployment of large-scale multimodal models.
• Deep understanding of generative architectures (diffusion, autoregressive, cross-modal).
• Expertise in scalable data curation and model pipeline optimization.

Responsibilities

• Lead research on pre-training and mid-training of multimodal foundation models at scale.
• Design and prototype novel architectures for real-time multimodal synthesis across modalities.
• Develop scalable data pipelines and training strategies for diverse datasets.
• Advance state-of-the-art techniques and bring research into production with engineering teams.
• Publish work in top-tier conferences and journals and communicate findings clearly.

Skills

Multimodal ML

Python

PyTorch

TensorFlow

Research leadership

Tools

Distributed training

GPU clusters

Job description

Pika is seeking a Research Scientist to lead the development of large-scale multimodal foundation models, focusing on pre-training and mid-training. You will design novel algorithms, curate data, and drive production-ready implementations in collaboration with engineering and product teams.

Ideal candidates have 5+ years of research experience, strong publication records, and hands-on expertise with PyTorch/TensorFlow.