Distillation Lead – Model Compression for Autonomous AI

About the role

Waabi is a leader in Physical AI, building the technology behind autonomous trucks and robotaxis. As the Distillation Lead you will own the strategy and execution for model distillation across Waabi’s AI stack, ensuring high‑performing models run efficiently in both onboard vehicles and large‑scale simulation pipelines.

Key responsibilities

• Define and drive the technical strategy for model distillation and compression across perception, world models, and planning.
• Design, implement, and scale state‑of‑the‑art distillation pipelines, including generative model distillation, quantization‑aware training, knowledge distillation, pruning, low‑rank factorization, and speculative decoding.
• Collaborate with ML Platform, Infrastructure, Onboard Autonomy, and Simulation teams to integrate compressed models into production, meeting latency, memory, and throughput targets.
• Define rigorous benchmarks and evaluation frameworks to assess efficiency vs. quality trade‑offs across hardware targets.
• Mentor researchers and engineers, champion best practices, and contribute to publications and open‑source projects.

Required profile

• Extensive hands‑on experience with model distillation, quantization, pruning, and compression for large‑scale neural networks in production.
• Strong research and engineering foundation; Bachelor’s or Master’s in ML, Computer Vision, Robotics, or equivalent industry experience.
• Proven technical leadership, setting direction and driving projects from concept to deployment.
• Experience collaborating with infrastructure, platform, and autonomy teams under real engineering constraints.
• Excellent communication skills to convey complex trade‑offs to diverse audiences.

Required skills

• Python
• PyTorch (or JAX)
• Distributed training at large scale
• TensorRT, ONNX, custom CUDA kernels
• Model distillation, quantization‑aware training, pruning, low‑rank factorization