Senior / Staff ML Systems Engineer – Framework Integration

Your mission

• Own the path from a model in a ML framework down to running fast on a target hardware. Work at all levels of the software stack with a wide understanding of everything from quantization to compiler abstractions to final target code generation.
• Bring up d reference models using Roofline's SDK, pinpoint which component an issue really lives in, and make sure it’s fixed at the root.
• Set technical direction across multiple teams for ML, compiler, runtime, and architecture based on your first-hand data . Define contracts and interfaces between the layers and keep the teams aligned to them.
• Turn your findings into products. Reflect on recurring blockers to craft solutions that benefit our customers and help improve our SDK. Think ahead to make sure that we always support the latest models and applications.
• Interacting with open-source community in the ML space (e.g. PyTorch)
• Raise the level of the engineers around you. Mentor junior engineers and help to maintain the best technical culture for the company.

Your profile

• 8+ years of hands-on experience in AI and embedded systems.
• Degree in Computer Science, Electrical Engineering, or related fields.
• Deep practical expertise in C/C++ and Python
• End-to-end experience with an accelerator toolchain for NPUs, DSPs, or GPUs spanning their compiler, runtime, kernel libraries and memory models. Strong communication and collaboration skills to align other engineers through clear ideas, abstractions, and design docs.
• Demonstrated technical influence beyond your own deliverables. Have product changes you argued for and got shipped, engineers you've mentored, or practices a team adopted because you established them.
• Ambition and curiosity to solve complex problems, a strong mindset to take ownership and drive our solutions forwards, and a collaborative attitude to foster our culture.

Nice to haves:

• Experience working with hardware or architecture teams, and an understanding of how their approaches to problems and processes differ from software teams’.
• Experience working with PyTorch, ONNX, Tensorflow
• Experience shipping an accelerated model into production under real constraints for applications such as automotive, medical, industrial or consumer where memory budgets, certification and time budgets apply.

Wir zeigen einen Auszug und einen Link je Anzeige, und den vollen Text, wenn die Quelle es erlaubt. Genannte Marken und Logos gehören ihren Inhabern.

Quelle: Statistisches Bundesamt (Destatis), Tabelle; dl-de/by-2-0; eigene Berechnung