AI Performance Engineer

ParTec AG is a fast-growing, agile technology and consulting company and a leading provider of modular high-performance computing systems. In addition to building complete computer systems, we offer our customers and partners software solutions, support, and consulting services. Our ambition is to become Europe’s leading integrator of quantum computers and to enable our customers to grow efficiently, cost-effectively, and continuously in an ever faster changing world and to adapt to future challenges and technologies.

ParTec seeks motivated engineers who will consistently deliver the high-quality results demanded by ParTec’s HPC and QC customers. Independent thinking and innovative approaches to address work activities and resolve issues is required. ParTec emphasizes a constructive, collaborative, and results-oriented approach to work.

For strengthening our technical pre-sales team for High-performance computing and Quantum Computing, we are looking for an

AI Performance Engineer

Job Summary

This position involves working with existing and emerging AI/ML workflows, to measure and optimise their performance on GPU and CPU architectures, as well as projecting performance to future systems.

Primary duties and responsibilities

Complete performance-based benchmark tests in a timely manner, often under short deadlines, to support customer workloads as well as internal product decisions.
Continuously monitor and analyze performance metrics to identify trends, patterns, and areas of improvement.
Generate reports and provide insights on performance to the management (experience with MS Office tools is required).
Understand modern AI fundamentals (neural net types, optimisation techniques, transformers), as well as performance estimation methodologies used to provide required information, and performance assessments for benchmarks on future ParTec and competitive systems.
Recommend configurations and settings for different types of hardware and software to be used by ParTec customers.
Maintain current knowledge of competitors’ and ParTec’s products and relevant AI optimization techniques to ensure ParTec’s ability to provide high-quality system and application performance results.
Collaborate closely with colleagues across the company, as well as peers outside the company, to produce optimal performance results of AI workflows run on ParTec platforms.
Building, debugging, running, optimizing and benchmarking popular existing and emerging AI workflows, to benchmark scale, quality and speed, as well as application areas such as foundational models, text and image classification or large language models.

Required skills and experience

Bachelor degree (Masters or PhD preferred) in Computer Science, Maths, Engineering, Physics, Chemistry, or similar.
Experience of AI benchmarking and performance optimisation.
The ability to convey highly technical materials to a variety of audiences both inside and outside of ParTec.
Keen analytical skills and problem-solving ability.
Desire and ability to work with leading edge technology.
Ability to work under demanding deadlines.
Ability to coordinate multiple tasks in a fast-paced challenging environment.
Fluency (written & spoken) in English is required.

Desired skills and experience

Extensive knowledge of Python and C++
Good knowledge of linear algebra
Experience with AI training and inference optimisations (pipeline, tensor, expert and data parallelism).
Experience with CUDA programming
Experience with non-CUDA GPU programming languages/APIs (HIP, ROCm, oneAPI)
Experience with TensorFlow, PyTotch or JAX
Experience with DeepSpeed, Horovod, HuggingFace transformers library, and/or other modern performance-oriented AI libraries
Experience with various GPU architectures (AMD, Intel, Nvidia)
Experience with MPI, OpenMP, nccl, rccl, gdr or gds libraries and optimisation
Experience with AI specific performance tools (profilers, debuggers)
Experience with CUDA or HIP code optimisation
Familiarity with Linux and shell scripting.
Familiarity with using HPC systems.
Desire and ability to quickly learn new skills.

Job conditions

Flexible work hours, flexible workplace including home office
Office environment available at our German headquarter in Munich
Occasional travels, mostly to destinations in Europe

This is an exciting opportunity in one of the most advanced and dynamically developing fields of computing.

If you are interested in working with us, please send your application with resume and salary expectations to career.sales@par-tec.com.