In May 2026, I took Sun Shines Bright to the REAL ML workshop in Johannesburg. Thanks to Namita Aavriti for encouraging me to apply.

These are notes on what I brought, what I asked for and who I was paired with.

What I brought

The question I’ve been working with:

Can community-learned, solar-powered small language models create new forms of technological intimacy that resist extraction while fostering care?

Where the project is:

• Early experiments in Jan-Mar 2025, then a workshop at NYU Shanghai in April 2025.
• Tested in Udaipur, India.
• Since September 2025, I’ve been training small language models on non-white feminist data.
• A small fund from New Media Design, HEAD Geneve, has supported research into data gaps in language models and collecting data.
• Workshops and tutorials so anyone can follow along.

The geographic focus is South Asia, mainly India.

What I went in looking for

I run this practice independently. I don’t have peers I can develop this work with, deeply and critically.

I was hoping for:

• guidance and collaboration on datasets for the fine-tuned models, including references for critical feminist writing and practice
• peer conversations on dataset collection: bias, ethics, labour, and other ways of collecting data
• honest feedback from people working at community and field level, on if and how this could be useful to them

What is next for my project

• Energy validation of the project
  • manufacturing
  • running inferences
  • is it really working?
• Rigorous documentation

My asks

• Collaborators who want to make their own funky, strange, lovable model, for them and their community. For now, only in English.