Work on generative audio systems across models, evaluation, and data
Design experiments that separate genuine progress from noise
Build evaluation and dataset pipelines that make model quality measurable and iteration faster
Make sound trade-offs across quality, latency, reliability, and cost
Comfort taking ownership in ambiguous problem spaces and staying engaged with the problem until it is solved
Genuine interest in audio, music, and generative modeling
Strong habits around evaluation, reproducibility, and performance
Fluency in Python and PyTorch, or similar tools
Generative modeling, including diffusion, autoregressive methods, or hybrids
Audio ML, or adjacent experience that transfers well, such as image generation
Multi-GPU or distributed training
High ownership over important technical work
Be at the forefront of AI-driven music innovation
Opportunity to work on infrastructure at scale
Competitive compensation and equity
Flexibility in how you work
Skills Required
- Ownership in ambiguous problem spaces
- Interest in audio, music, and generative modeling
- Strong habits around evaluation, reproducibility, and performance
- Fluency in Python and PyTorch or similar tools
- Experience with generative modeling (diffusion, autoregressive, or hybrids)
- Audio ML or closely related generative modeling experience (e.g., image generation)
- Multi-GPU or distributed training experience
What We Do
GRAI is an AI music research lab based in Warsaw, Poland, dedicated to building foundation models and interaction primitives for the future of music. The company develops AI-powered tools that enable users to interactively remix, transform, and share songs, aiming to make music more social and interactive while ensuring artists maintain control and potentially benefit from new royalty streams.
