AI Systems Researcher | Machine Learning Systems, Compilers & Efficient Inference

Ailven LIU 刘翔

AI Systems Researcher, sudo robotics

Ailven Liu is an AI Systems Researcher at sudo robotics · Foundation Models, working on model–system co-design for embodied foundation models. Their work connects model design, compiler abstractions and inference behaviour, integrating computational efficiency into the design of the models themselves.

Previously, Ailven was an AI Infrastructure R&D Engineer at NIO · Autonomous Driving, developing production AI infrastructure across model deployment, compilation and runtime systems. This experience provides practical grounding in how models operate within the performance and resource constraints of real computing environments. Model deployment involves adapting computation to the target platform’s execution model while preserving intended model behaviour.

Ailven’s research in model compression and multimodal generation examines how model representations shape capability, computational cost and practical deployment. It connects algorithm design with compilation and inference, treating efficiency as a joint property of models and their execution systems. Compression examines which information must be retained for task quality and how representations can be simplified to reduce computational cost. In multimodal generation, representations must support coherence across modalities; their form also affects the computation required to produce an output.