Research Engineer / Scientist, Foundational Multimodal Models

About Fixie We’re a Seattle-based AI startup (with support for working remotely). We’ve raised $17M in seed funding. Our vision is simple: build artificial intelligences that can communicate as naturally as humans. We’re a small team of researchers and engineers with a deep focus in speech and real-time technologies. Our core model, Ultravox, is open-source. We also build a serving stack that’s optimized for very low-latency interactions. The Role As a Research Engineer & Scientist, you will lead the effort to develop next-generation foundational multimodal models that power Ultravox, our open-source speech-to-speech model. What you’ll do

• Lead critical research on architectural design, pre-training, and post-training of foundational multimodal models to develop real-time voice AI.

• Collaborate with a team of researchers and engineers to develop foundational multimodal models with comprehensive capabilities in speech understanding, speech generation, and full-duplex real-time communication.

• Develop novel models based on public and proprietary data sources.

• Build tools to improve our data flywheel and measure model quality.

• Drive the optimization and deployment of AI models for real-world applications in partnership with engineering and product teams. Things we’re looking for

• An incredibly strong AI researcher with a track record of contributions to AI research, systems, and products.

• Experience with large language models, speech models, and multimodal models.

• Strong experience in Python and, ideally, PyTorch.

• Ability to roll up your sleeves and get things done.

• A great communicator and team player. Benefits

• Generous equity package

• Unlimited PTO (take time when you need it)

• Top-of-market salary

• Great healthcare

• 401k with match Apply