We build multi-scale, high-fidelity, scalable brain-simulation platforms and computational models. From foundational software and parallel algorithms to modeling theory and computational models, we push end-to-end simulation from single neurons through circuits to the whole brain.
Our long-term goal is a biologically realistic, functionally meaningful whole-brain model — and the software, algorithms, and infrastructure to scale brain-dynamics simulation to hundreds of billions of neurons — applied to decoding brain function, studying disease mechanisms, and powering next-generation brain-inspired AI.
BrainX Software Ecosystem
We are building the BrainX brain-simulation ecosystem — a multi-scale modeling toolchain spanning ion channels, cells, circuits, and whole-brain networks — with efficient parallel algorithms and modern software architectures for distributed brain modeling, targeting CPUs, GPUs, TPUs, and heterogeneous systems, and scaling whole-brain dynamics to hundreds of billions of neurons.
BrainPy 面向脑动力学建模的通用编程框架。
BrainTrace 基于资格迹的脑动力学在线学习工具。
BrainUnit 面向脑动力学的物理单位和单位感知数学系统。
BrainCell 面向生物细节脑细胞建模的高效仿真工具。
BrainMass 面向大尺度脑动力学的神经质量模型工具。
BrainState 面向 CPU、GPU 和 TPU 的状态变换与高效仿真系统。
BrainEvent 面向脑动力学的事件驱动计算工具。
BrainTaichi 面向脑动力学仿真的 Taichi 加速后端。
BrainTools 脑动力学编程通用工具集。
Modeling Algorithms
Whole-brain simulation must balance biological realism, computational efficiency, interpretability, and learning capacity. Building on BrainTrace, we develop resource-efficient algorithms, online learning for spiking networks, modular cognitive learning mechanisms, and data- and task-driven multi-scale modeling — connecting structure, dynamics, and cognition.
Computational Models
Understanding the brain requires connecting molecular, cellular, circuit, and systems-level data. We combine simulation software, algorithms, and multi-modal data to build hierarchical brain-dynamics models for neural dynamics, cognition, disease mechanisms, and biologically compatible AI.
