GSoC 2026: Interested in Project 4 – DataLoader Support for Large-Scale Training

[rsvr76]

Hi everyone,

I'm Ram (GitHub: rsvr76), a Python/ML developer interested in Project 4 for GSoC 2026.

I came across sbi while looking at the GSoC project list and Project 4 immediately stood out.
I've spent a lot of time working with PyTorch DataLoaders in ML competitions so the problem space feels very familiar to me.

I already made a small contribution to get started - I worked on issue #1035 (BatchNorm → LayerNorm) and submitted a PR for it.
A couple of things I wanted to ask before writing the proposal:

• What's the biggest pain point right now with large-scale training in sbi? I want to make sure I'm solving the right problem

• Any specific areas of the codebase I should focus on to understand the current data pipeline better?

I've also joined the Discord. Happy to discuss further there if that's easier.

[janfb]

Hi @rsvr76 thanks for your interested in applying for GSoC with us!

Please join the Discord Server if you haven't already and reach out to me directly to discuss next steps.

Cheers,
Jan

[rsvr76]

Thanks @janfb I've joined the Discord and will reach out to you there directly.