Spotify - Staff Research Engineer

• Close Collaboration: Work side-by-side with research scientists to conduct ground breaking research in music generation (diffusion, flow matching, or autoregressive models), as well as related domains like ML-based audio processing, music information retrieval, machine learning, and signal processing. • Improve model training pipelines. You’ll debug distributed training, optimize data loading at massive scale, and ensure smooth scaling across compute environments. • Optimize performance. You’ll profile and accelerate existing training and inference code to make experiments faster and production systems more responsive. • Integrate models into production environments. You’ll work directly with platform and product teams to deploy models into the hands of hundreds of millions of Spotify’s users. • Incorporate state-of-the-art research. You'll translate models and techniques described in the literature into robust, well-engineered prototypes. • Maintain a high-quality codebase. You’ll enforce clear structure, consistency, and testing practices to support long-term maintainability on a codebase shared between members of a fast-paced globally distributed team. • Enhance researcher experience. You’ll build internal tooling, libraries, and workflows to make experimentation, debugging, and deployment more efficient for the whole team. • You have experience training or fine-tuning large machine learning models on GPUs using PyTorch or similar frameworks. • You have experience working with cloud platforms like Google Cloud Platform, AWS, or Microsoft Azure. • You understand how to debug problems in machine learning training code. • You communicate effectively with global teams and are ready to work both face-to-face and asynchronously with collaborators on multiple continents. • You have experience optimizing code for performance and can make GPUs “go brrr” (train at maximum efficiency). • You learn new concepts and technologies quickly and keep up to date with the rapid pace of development in machine learning and AI. • You are resourceful and proactive; when faced with blockers, you seek out solutions through research, experimentation, and collaboration. • You’re not afraid to dig deep into the stack: working with lldb, NVIDIA Nsight, or other low-level debugging tools is a plus. • You have a solid grasp of computer science concepts like type systems, compilers, parallelism, thread safety, encapsulation, and the like. • You have an interest in learning more about audio processing and music information retrieval and you're excited about building amazing products that use these technologies. • We offer you the flexibility to work where you work best! For this role, you can be within the EMEA region as long as we have a work location • This team operates within the Central European and GMT time zone for collaboration. • Core working hours are CET 3pm-6pm / EST 9am-12pm.