RESEARCHER, EFFICIENT INFERENCE

What you'll do

• Develop and evaluate quantization and speculative decoding methods, including training-time techniques that improve inference efficiency.
• Run controlled experiments on real workloads and move promising research into production with inference engineers.
• Review research literature, publish or share findings, and coordinate efficiency choices with model and training researchers.

What they ask for

• Strong ML research track record in efficiency methods such as quantization, speculative decoding, distillation, MoE, or sparse attention.
• Deep familiarity with training and inference performance; experience moving efficiency research from prototype to production.
• Fluency in PyTorch, Jax or equivalent, with comfort at the kernel and serving-framework level.
• Strong statistical expertise and written communication; published research at NeurIPS, ICML, ICLR, MLSys, or comparable venues.

Nice to have

• PhD in ML, systems, or a related field.
• Open-source contributions to efficient-inference libraries, hardware-aware optimization experience, or background in numerical methods and low-precision or approximate computation.

Tools

PyTorch, Jax

AI focus

Researches methods that make machine learning models more efficient to train and run, including quantization, speculative decoding, sparse attention, distillation, and mixture-of-experts inference.