Hello everyone,
My name is Akshit Rai and I am currently a 2nd year Computer Science and Artificial Intelligence student at Vishwakarma Institute of Technology, Pune. I have experience working with Python, C++, and machine learning. In the past I have worked with various neural network architectures such as CNNs, RNNs, U-Nets, and briefly GNNs. Recently I have also started exploring spiking neural network simulators and neuroscience-related tools.
I am very interested in the GSoC 2026 project on improving serialization and deserialization support for Brian and brian2tools. I have started going through the Brian2 and brian2tools repositories to understand the current mechanisms used for storing simulation state and exporting models.
From what I have seen so far, the current store/restore mechanism mainly handles in-memory snapshots of simulation state variables, but there seems to be limited support for exporting a complete network architecture or restoring simulations across different backends such as Python runtime and C++ standalone mode. One possible direction I am exploring is using a backend-independent representation of the network architecture and state, combined with an efficient storage format such as HDF5 for large numerical data (e.g., connectivity structures and state variables).
Over the next few days I plan to:
• Explore the Brian2 and brian2tools repositories in more detail
• Understand how the store/restore system works internally
• Study the existing basicexporter implementation
• Experiment with exporting a small network representation
Since the proposal timeline is approaching, I am currently trying to quickly familiarize myself with the relevant parts of the codebase.
If there are particular parts of the codebase, design discussions, or documentation that would be especially useful for understanding the current serialization approach, I would greatly appreciate any guidance.
Also, are there any recommended prerequisite tasks or small issues that prospective contributors should look at before working on a proposal for this project?
@mstimberg @d.goodman
Looking forward to learning more and contributing.
Best regards,
Akshit Rai