Lead Architect, Multimodal Transformers

Omaze; Mountain View (CA); On-site; USD 200,000 - 230,000; Full time

Benefits offered by this job

Competitive compensation

Equity bonus

Excellent health plan

Dental and Vision

Flexible PTO

Office in Mountain View

Commuter benefits

Retirement plan

Job summary

Kodiak Robotics, Inc. in Mountain View, CA is seeking engineers to advance AI-powered autonomous trucking.

The role focuses on designing multimodal transformers that fuse camera, LiDAR, and radar data into cohesive representations, with emphasis on cross-attention and scalable training across large datasets. Ideal candidates have 4+ years in transformer architectures, Python proficiency, and hands-on experience with PyTorch or TensorFlow.

Qualifications

• BS, MS, or PhD in AI, Computer Science, or a related field.
• 4+ years experience with transformer architectures, particularly in multimodal or multi-stream settings.
• Familiarity with cross-attention, token fusion, or modality alignment techniques.
• Proficiency in Python and deep learning frameworks like PyTorch or TensorFlow.
• Strong understanding of scalable training for large models, including distributed training and mixed-precision optimization.

Responsibilities

• Design and develop multimodal transformer architectures that fuse camera, LiDAR, and radar into unified representations
• Research and implement cross-modal attention mechanisms, token fusion strategies, and efficient multi-stream tokenization
• Build scalable training pipelines for large-scale multimodal transformers across massive real-world datasets
• Explore self-supervised and contrastive pretraining objectives that learn transferable multimodal representations
• Optimize transformer models for real-time inference under latency and compute constraints

Skills

Transformer architectures

Multimodal learning

Cross-modal attention

Python

Education

BS/MS/PhD in AI or Computer Science

Tools

PyTorch

TensorFlow

Job description

Kodiak Robotics, Inc. in Mountain View, CA is seeking engineers to advance AI-powered autonomous trucking.

The role focuses on designing multimodal transformers that fuse camera, LiDAR, and radar data into cohesive representations, with emphasis on cross-attention and scalable training across large datasets. Ideal candidates have 4+ years in transformer architectures, Python proficiency, and hands-on experience with PyTorch or TensorFlow.