Weitong Qian钱韦潼

An open-source agent system that autonomously writes, profiles, and optimizes CUDA kernels — pushing AI further down its own software stack.

Combined Algorithm Selection and Hyperparameter optimization (CASH) with an LLM and Bayesian optimization sharing one Monte Carlo search tree. The LLM reads the search path and proposes semantically informed candidates; Bayesian optimization fits algorithm-specific surrogates; a reliability-aware policy hands control from the LLM to BO as the surrogates improve. Outperforms BO, LLM, and hybrid baselines on 104 AMLB datasets.

A memory-centric agent system for the full scientific research lifecycle — from literature ingestion and idea generation to experiment execution, manuscript writing, and reviewer rebuttal. Structured persistent memory separates reusable knowledge from project-level artifacts; a feedback-driven self-improvement loop converts experiment outcomes and review signals into versioned updates to skills and memory, so the system compounds across projects.

An AI research-onboarding tool, built on the OpenClaw skill platform. Given a topic, it constructs a self-contained, growing knowledge base — field overview, foundational and frontier papers, peer reviews, knowledge graph — that a newcomer can converse with directly.

Vision-based, RL-driven non-prehensile grasping; RL controllers for stable and adaptive locomotion on quadruped and biped robots over varied terrain. Advised by Prof. He Wang.

A system in which embodied AI agents autonomously plan and conduct scientific experiments — research automation extended into the physical world.