MiddleOfficeLondon
Neural Network Performance Engineer
О роли
Описание вакансии
About the company
- Humanoid is building the world's most capable, commercially-scalable, and safe humanoid robots.
- HMND-01 Alpha is running in real industrial pilots.
Responsibilities
- Analyze performance bottlenecks of model architectures and come up with potential improvements.
- Make the model run on new hardware (e.g. NVIDIA Thor) efficiently.
- Implement custom kernels to reduce memory throughput requirements.
- Quantize a model with minimal loss of quality.
- Suggest and implement changes of model architecture that will enable better performance characteristics.
Requirements
- 3+ years building deep-learning systems (industry or research) with shipped models or published artifacts.
- 1+ years experience working on performance of neural network inference.
- Excellent understanding of GPU architecture.
- Strong Python + PyTorch/JAX; profiling, debugging numerics, writing maintainable research code.
- Clear documentation and communication of trade-offs.
Nice to have
- Robotics or autonomous driving experience.
- Open source code showcasing inference performance improvements.
- Publications at ICLR/ICML/NeurIPS or equivalent.
- Familiarity with vision-language (VLM) or vision-language-action (VLA) models.
Conditions
- Competitive equity: stock options.
- 30+ paid days off, including 23 days of annual leave, all UK bank holidays, and additional company closure days.
- Private healthcare, including virtual and in-person care.
- Pension scheme with 8% total contribution (5% employee, 3% employer).
- Free daily breakfast, catered lunch, and snacks in-office.
Стек и навыки
С чем работаем