ifm Benefits Overview

Develop and optimize CUDA kernels and machine learning software for training and inference on advanced GPUs. Profile workloads, improve GPU utilization, build benchmarking and automation tools, troubleshoot system bottlenecks, support new model architectures, validate numerical and gradient correctness, review code, and contribute to technical documentation and industry events.

Develop and scale distributed machine learning infrastructure for large-scale LLM pre-training. Responsibilities include extending frameworks such as DeepSpeed and FSDP, implementing distributed optimizers, managing multi-node GPU launches, building experiment tracking and monitoring, improving reliability and performance, and maintaining data-loading and checkpointing workflows. The role requires strong Python, systems engineering, distributed training, multi-node debugging, and large-scale pre-training experience.

Build and scale end-to-end reinforcement learning infrastructure for foundation models. Responsibilities include extending distributed training frameworks, integrating rollout and inference systems, implementing reward and trajectory processing, configuring multi-node GPU launches, monitoring experiments, optimizing performance, and developing checkpointing and failure-recovery systems. The role requires strong Python and ML systems engineering experience, hands-on LLM RL pipeline ownership, and expertise with distributed training and policy optimization methods.