Over deze vacature
The Machine Learning Systems Engineer contributes to ML infrastructure and the open-source ScalarLM codebase.
Remote (San Francisco Bay Area / North or South America) 3+ years in ML engineering or research
The Machine Learning Systems Engineer contributes to ML infrastructure and the open-source ScalarLM codebase. The role intersects high-performance computing, distributed systems, and advanced machine learning research.
Develop and optimise distributed training algorithms for large language models (LLMs). Implement high-performance inference engines and optimisation techniques. Work on integrations between the vLLM, Megatron-LM and HuggingFace ecosystems. Build tools for seamless model training, fine-tuning and deployment. Optimise performance across advanced GPU architectures. Research and implement new techniques for self-improving AI agents.
Proficiency in both C/C++ and Python. Deep understanding of HPC concepts, including MPI programming and distributed computing across multiple GPUs/nodes. Experience with transformer architectures and distributed training techniques (data parallelism, model parallelism). Experience with large-scale distributed training frameworks (Megatron-LM, DeepSpeed) and inference-optimisation frameworks (vLLM, TensorRT).
BlueStone Solutions B.V. is certified in accordance with NEN 4400-1 and recognised sponsor with the IND.