Collaborative & Private AI Model Training Use-Cases

Project description:
Exploring how mutually distrusting parties can collaboratively and privately train AI models, with applications in community-driven, responsible, and privacy-preserving contexts.

Artifacts:

• Slides: Collaborative and private AI model training to combat algorithmic bias
• Github

Contributors:
Yuriko Nishijima, Val Elefante

Demo Day Video(s):

Val on use cases research (midpoint demos):

Yuriko on technical implementations (midpoint demos):

Val (final demo day):

Support requested from the community:

• Mentorship and expertise
• Connections to funding sources
• Help publicizing the work
  Yuriko is looking for a funding resource to continue this research!

Further comments:

Yuriko:
I’m an independent applied cryptography researcher with expertise in programmable cryptography (ZK, FHE, MPC, etc.) and Ethereum development. Currently, I’m exploring how cryptography and blockchain can support machine learning — not just for privacy protection, but to enable mutually distrusting parties to train models collaboratively with fair data contribution rewards (see: “data dignity”).
Find me on Telegram/Discord: @yuriko627, Twitter: @yurikonishijima

Val:
I am an independent researcher and project manager focused on community-led, responsible, public interest tech development. Like Yuriko, I’m excited about private and collaborative AI model training for mission- and community-driven orgs. Potential partners include CDFIs (credit unions and banks serving marginalized groups), social movement data co-ops, and artist collectives.
Reach me on Discord: @yogival, Signal: velefante22.82, Twitter: @velefante22