4 months ago
London, United KingdomMid Level
Responsibilities
- Analyze model-architecture performance bottlenecks and identify potential improvements.
- Optimize neural-network models for new hardware, including NVIDIA Thor.
- Implement custom kernels to reduce memory-throughput requirements.
- Quantize models while minimizing quality loss.
- Design and implement model-architecture changes that improve performance without sacrificing capability.
- Profile and debug neural-network inference performance and numerical behavior.
Requirements
- At least 3 years of experience building deep-learning systems in industry or research, with shipped models or published artifacts.
- At least 1 year of experience optimizing neural-network inference, including bottleneck analysis, custom kernels, quantization, and deep-learning compilers.
- Excellent understanding of GPU architecture and model-performance characteristics.
- Strong Python and PyTorch or JAX skills, including profiling, numerical debugging, and maintainable research code.
- Ability to document experiments clearly and communicate trade-offs effectively.
- Preferred: robotics or autonomous-driving experience; open-source code demonstrating inference-performance improvements; ICLR, ICML, NeurIPS, or equivalent publications or contributions; and familiarity with VLM or VLA models.
Benefits
- Competitive equity through stock options.
- 30+ paid days off, including annual leave, UK bank holidays, and additional company closure days.
- Private healthcare with virtual and in-person care.
- Pension scheme with an 8% total contribution, consisting of 5% employee and 3% employer contributions.
- Free daily breakfast, catered lunch, and in-office snacks.
- Opportunity to work with engineers, researchers, and product experts on humanoid robotics and AI.
- Significant ownership, access to founding leadership, and influence over product direction from day one.
- Based in London and working in-office.
Categories
