Machine Learning Engineer – Multimodal Modeling | San Francisco (Hybrid) | Up to $325k Base + E

Most machine learning roles today focus on building applications on top of foundation models.

This opportunity is different.

You'll join the Applied Science team of a rapidly scaling, venture-backed startup building a new class of AI-powered risk engine. Rather than relying on broad geographic averages like traditional insurers, the company combines physics-based digital twins, CFD and multiphysics simulations, and AI to model catastrophic risk at the individual property level - enabling a fundamentally different approach to underwriting. The goal isn't simply to interpret the world, it's to model it.

This is an opportunity to work on a class of machine learning problems that very few engineers ever get exposure to.

As a Machine Learning Engineer, you'll design and train multimodal model architectures, develop large-scale training and evaluation pipelines, and take models from architecture and experimentation through production deployment. You'll also contribute to agentic AI systems capable of interacting with complex workflows and real-world decision making.

You could be a strong fit if you've worked with:

• Vision-Language Models (VLMs) or multimodal foundation models
• Agentic AI, tool use, or reasoning systems
• Representation learning, retrieval, or embedding systems
• Robotics, simulation, medical imaging, remote sensing, molecular AI, digital twins, or other physics-informed machine learning domains

If you're interested in solving machine learning problems that push beyond traditional multimodal AI, I'd love to tell you more.