Researcher, Efficient Inference

About the Company

We're building autonomous research agents for recursive self-improvement (multi-agent systems that propose, run, and analyze machine learning experiments). We're a small team based in San Francisco.

Responsibilities

• Research and develop quantization methods including post-training quantization and quantization-aware training
• Design and evaluate speculative decoding approaches such as draft models and tree attention
• Investigate training-time efficiency methods like distillation, sparse attention, and mixture-of-experts
• Run controlled experiments at production scale to characterize performance on real workloads
• Co-design methods with the inference engineering team to ensure results are production-ready
• Partner with model and training researchers to align efficiency choices with model architecture

Requirements

• 5+ years of hands-on research experience
• Strong track record of ML research on efficiency methods (quantization, speculative decoding, distillation, MoE, or sparse attention)
• Deep familiarity with training and inference performance characteristics
• Proficiency in PyTorch, Jax, or equivalent
• Experience moving efficiency research from prototype to production
• Strong statistical expertise and written communication skills
• Published research at NeurIPS, ICML, ICLR, MLSys, or comparable venues

Preferred Qualifications

• PhD in ML, systems, or a related field
• Open-source contributions to quantization or efficient-inference libraries
• Experience with hardware-aware optimization and accelerator-specific tooling
• Background in numerical methods or low-precision arithmetic