Member of Technical Staff, ML Performance

Odyssey is pioneering world models, the next frontier of artificial intelligence. By learning from the real-world, Odyssey is training a new kind of generative model, capable of generating cinematic, interactive worlds in real-time. Odyssey’s mission is to reinvent film, gaming, and beyond.

Odyssey was founded in late 2023 by Oliver Cameron (Cruise, Voyage) and Jeff Hawke (Wayve, Oxford AI PhD), two veterans of self-driving cars and AI. They’ve since recruited a world-class team of AI researchers from Cruise, Waymo, Wayve, Tesla, Microsoft, Meta, and NVIDIA; lead computer graphics researchers from EA, Ubisoft, and Valve; and technical artists behind Hollywood blockbusters like Dune, Godzilla, Avengers, and Jurassic World.

Odyssey has raised significant venture capital from GV, EQT Ventures, Air Street Capital, DCVC, Elad Gil, Garry Tan, Soleio, Jeff Dean, Kyle Vogt, Qasar Younis, Guillermo Rauch, Soumith Chintala, and researchers from OpenAI, DeepMind, Meta, and Midjourney. Ed Catmull, the founder of Pixar, serves on Odyssey’s board.

The Role

We're seeking a talented engineer passionate about advancing AI models. We're building inference infrastructure to scale to hundreds of thousands of users within a year. Your focus will be ensuring our models deliver exceptional speed, reliability, and scalability while optimizing efficiency to minimize TFLOPS per user.

You will

• Optimize models that will be used in real-time by hundreds of thousands of users
• Partner with our elite team of ML researchers and engineers
• Develop sophisticated tools to identify performance bottlenecks and stability issues
• Pioneer innovative approaches, frameworks, and system designs that enhance performance metrics across our model inference infrastructure
• Have significant autonomy in technical decisions
• Use the latest-generation GPUs

Who You Are

Ideal Qualifications

• 8+ years of software engineering experience, with significant work in ML Performance
• Deep insight into modern machine learning architectures with a natural instinct for performance optimization, particularly inference
• Track record of owning projects end to end
• Problem-solving mindset with the ability to acquire new skills as needed
• Proficiency with PyTorch (or TF/JAX), as well as NVIDIA GPU ecosystems and optimization stacks
• Highly metric-based
• Strong Python and C++ skills

Bonus Qualifications

• Experience optimizing kernels with Triton or CUDA
• Enjoy completely reimagining and reconstructing production systems
• Experience with large models (>100M parameters)
• You think a peregrine falcon is slow