Next Brain Lab · Independent research

Simulate neural systems.
Discover mechanisms.
Build memory.

We study how neural structure and dynamics give rise to learning and memory, and turn those insights into computational tools and AI systems.

An independent research lab at NEXT BRAIN LTD, founded and led by Xinjun (Rock) Li. Fundamental research and practical development advance together.

Our research Work with us

Research programme

Three independent directions.
Room to work together.

Each direction has its own scientific questions and outputs. Together, they connect the tools we build with the mechanisms we want to understand.

Neural Simulation

Foundation available · Brian2 Atlas

How can we simulate neural systems faithfully, at scale, across computing architectures?

We build the computational tools needed to investigate neural dynamics. Brian2 Atlas is our current foundation; scaling, execution fidelity and hardware mapping are research questions in their own right.

Reproducible execution across CPU, GPU and distributed systems
Large-scale networks and the trade-offs between fidelity, memory and compute
Future compilation and mapping to neuromorphic hardware

Neural Circuits & Motifs

Research agenda

What are the smallest functional circuits that can learn, adapt and support evolution?

We seek compact mechanisms whose function can be tested through intervention. The focus is on how connectivity, dynamics and plasticity work together, and how useful motifs can evolve and combine.

Minimal circuits with measurable learning and adaptive behaviour
Evolutionary search, robustness and the composition of motifs
Causal tests through ablation, perturbation and matched controls

Memory Mechanisms

Research agenda

How can neural systems store, retrieve and update useful information over time?

We study memory as a dynamic process, from local plasticity to persistent network state. We aim to translate testable neural mechanisms into memory systems for artificial intelligence.

Encoding, retrieval, consolidation and forgetting across timescales
Spiking and conventional neural networks with persistent, adaptable memory
SNN-inspired memory modules for language models and continual learning

A shared approach

NeuroAI connects
the programme.

NeuroAI runs through all three directions. We use neuroscience to formulate computational hypotheses, and AI methods to search, model and test them.

In simulation, this means tools for studying biological and artificial networks. In motifs, it means discovering reusable learning mechanisms. In memory, it means testing whether neural principles can improve artificial systems, including conventional neural networks and language models.

Every proposed connection needs a measurable experiment: a mechanism, a baseline and a test of what changes.

Current work

Start with the evidence.

Atlas is the foundation we have released. Motif discovery and memory systems define the next research agenda.

Open source · Available on arXiv

Brian2 Atlas

An alternative simulation backend that retains Brian 2 as the modelling frontend, with its own intermediate representation, code generation and Rust execution stack. Supported models can run across CPU, GPU and distributed execution paths.

The overview documents the architecture, supported scope, experiments and limitations.

Explore the engine Source code Read on arXiv Results & evidence

Research that can be examined

We share methods, code and results with their assumptions and limitations. Reproduction, meaningful controls and negative results help determine what to pursue next.

From mechanisms to useful systems

Research informs the Next Brain platform and future applications. Practical constraints, in turn, help us ask sharper questions about efficiency, robustness and learning.

Collaborate with the Lab

Bring a question
we can test together.

We welcome conversations with academic groups, independent researchers and research engineers working on simulation, neural circuits or memory.

PhD researchers interested in part-time or project-based research are also welcome to get in touch. Tell us about your interests, relevant work and a question you would like to investigate.

Contact the Lab About Next Brain

[email protected]