We're seeking exceptional Research Scientists to advance the state-of-the-art in multimodal AI systems deployed in real-world environments. This role sits at the intersection of cutting-edge research and practical applications, focusing on the entire model lifecycle from pre-training to deployment and continuous improvement.

The Opportunity

As a Research Scientist on our team, you'll work on frontier multimodal models capable of understanding and reasoning across diverse input modalities including text, images, audio, and sensor data. Your research will directly impact how these systems perform in complex, real-world settings with real users.

Your work will span:

• Model Architecture & Pre-training: Design and implement novel architectures, algorithms and systems for efficiently processing multimodal inputs at scale
• RL & Post-training Techniques: Novel RL and post-training infrastructure and algorithms to scale multimodal test-time compute
• Inference Aware Design: Research efficient architectures, streaming systems, model compression, distillation, quantization.
• Evaluation & Benchmarking: Create rigorous evaluation frameworks that assess real-world performance beyond standard benchmarks

What You'll Bring

• BS, MS, PhD in Machine Learning, Computer Science, or related field (or equivalent research experience)
• Strong publication record in multimodal learning, transfer learning, or deployment of ML systems
• Experience with modern deep learning frameworks (PyTorch, JAX, TensorFlow)
• Track record of implementing and scaling large neural network architectures
• Ability to translate research insights into practical engineering solutions
• Excellent communication skills to collaborate with cross-functional teams

Ideal Candidates Will Also Have

• Experience deploying ML models in production environments
• Familiarity with model optimization techniques (pruning, quantization, distillation)
• Background in systems that integrate multiple modalities (vision-language, audio-visual, etc.)
• Understanding of ML infrastructure and distributed training