Senior Distributed ML Engineer

Kurz zusammengefasst

The role of Senior Distributed ML Engineer at LawZero involves working on a novel AI safety agenda. LawZero is a non-profit organization dedicated to advancing research and creating technical solutions for safe-by-design AI systems. The company focuses on developing AI that understands the world rather than acting in it, aiming to accelerate scientific discovery and provide oversight for other AI systems. The main day-to-day responsibilities include collaborating with ML research scientists to solve complex training and inference problems using large models. This involves investigating performance bottlenecks, profiling research experiment code, and optimizing the utilization of computing resources. Additionally, the role requires developing tools and libraries to simplify the use of distributed computing resources and establishing best practices for large-scale, distributed ML model development workflows. To qualify for this role, candidates must have a degree in a relevant computer science field, such as computer science, computer engineering, or software engineering. A minimum of three years of experience in designing and implementing distributed ML training frameworks is required, with recent experience using tools like Megatron, DeepSpeed, HuggingFace Accelerate, FSDP, vLLM, and/or verl. Familiarity with cloud platforms, GPU profiling tools, and containerization tools is also essential. Nice-to-have skills include experience with workload managers, data infrastructures, and platforms. The role offers a collaborative and inclusive work environment, with a focus on diversity and respect for differences. The position is based in Montreal, where LawZero is committed to advancing AI as a global public good. What the role asks for: - A degree in a relevant computer science field. - 3+ years of experience in designing and implementing distributed ML training frameworks. - Recent experience using Megatron, DeepSpeed, HuggingFace Accelerate, FSDP, vLLM, and/or verl. - Ability to collaborate effectively with cross-functional teams. - Experience with cloud platforms (e.g., AWS, GCP, Azure). - Experience with workload managers (e.g., Ray, SLURM). - Experience with GPU profiling tools (e.g. PyTorch profiler, PyProf, NVIDIA Nsight). - Familiarity with containerization tools (e.g., gRPC, Docker, Kubernetes). - Familiarity with data infrastructures and platforms (e.g., vector databases).