AI Research Engineer (Kernel & Inference Optimization)

This full-time position, based in Germany, involves designing and deploying high-performance model-serving architectures for advanced AI systems. The day-to-day work centers on developing inference pipelines that function effectively across diverse hardware, including resource-constrained mobile and edge devices. You will focus on optimizing latency, throughput, memory efficiency, and scalability by creating novel inference strategies and GPU kernels. The role requires working with complex architectures such as text, image, audio, diffusion models, and vision transformers. Your responsibilities include establishing clear performance targets for response time, token generation speed, and memory footprint. You will build and execute controlled benchmarks in both simulated and production environments to track key metrics like error rates and consumption. Additionally, you must create representative datasets and simulation scenarios to evaluate model performance under real-world conditions. The job combines hands-on research with low-level engineering to identify computational bottlenecks and translate findings into measurable improvements. The advert does not specify particular software tools, required qualifications, or language requirements. It also does not detail specific experience levels or benefits. The working arrangement is described as a highly technical, remote environment where you will collaborate with cross-functional teams. This role suits candidates interested in pushing the boundaries of efficient AI systems through rigorous testing and iterative optimization. Applications are managed by a partner company on behalf of the employer.