Grant submission for nymja.AI, your open-source LLM with built-in privacy

🪻 Nym Squad League Proposals
1

Grant submission for nymja.AI, your open-source LLM with built-in privacy.

General Information:

Nymja.AI is an app that integrates the power of open-source generative large language models with the security of the Nym mixnet architecture.

In a 6-month project, we will deliver an MVP, which allows the user to interact by text with Mistral’s LLM (Apache 2.0 license) running on top of Nym mixnet via a web application.

In a second stage of another 6 months (1/25-06/2025), it will be possible to extend the project to a production launch; with the service hosted remotely, having carried out intense batteries of tests and adding new features (such as audio and photo communication).

Nymja.AI is the solution that incorporates privacy into the most modern and powerful AI tools.

Team:

Development plan:

07/2024-08/2024:

  • Project conceptualization and feasibility, cost and software architecture research.

08/2024-09/2024:

  • Web page design in Figma.

09/2024-10/2024:

  • Front-end
  • LLM running locally

10/2024-11/2024:

  • LLM/front-end integration

11/2024-12/2024:

  • Mixnet + app integration

  • Testing and a local alfa version.

Costs and financial plan:

07-09/2024 (70k-100k Nym)

  • Developers and designers salary
    • 11k-16k for 3 months, for 2 professionals.
      • 66k-96k NYM.
    • Consultancy and Courses
      • 2.5k NYM
    • Miscellaneous Expenses:
      • 1.5k NYM

10-12/2024 (80-100k NYM)

  • Developers salary:
    • 11k-14.3k for 3 months, 2 professionals.
      • 66k-86k NYM
    • Cloud storage (the MVP will run locally, but we will already starts testing its web serving) - 4k NYM
    • Security Audition - 7k NYM
    • Miscellaneous Expenses:
      • 3k NYM.

Benefits for Nym and community:

Nym will have at its disposal the alpha version of Njmja.AI, AI chat application running on top of Nym mixnet infrastructure, uniting Nym’s and Mistral open-source philosophies.

It’s a tool with huge market potential, as it combines the finest advances in LLM with the world’s strongest privacy architecture.

The project’s scalability potential is significant. In addition, the app will also serve as a convenient tool for the Nym community itself to have a practical web AI alternative with built-in privacy. No more personal information tracking, obscure privacy policies and closed-source code.

Captura de Tela 2024-08-15 às 15.42.15
Captura de Tela 2024-08-15 às 15.42.151124×251 62.5 KB

2 Likes
2

Great proposal colleagues. It is a very interesting integration with a lot of potential. We will talk about it in the next NSO Squad meetings, in case we can contribute and support in any way the proposal. :fist: :brown_heart:

1 Like
3

We’d love feedback on the design and user experience as soon as we have the first stage ready :slight_smile:

1 Like
4

I’m not an AI specialist, but as far as I can see, the platform is paid, so users will have to pay a certain amount of money to continue using the product?

1 Like
5

They have 2 free licenses: Apache 2.0 (free for commercial use) and MNPL (free for research).

https://www.apache.org/licenses/LICENSE-2.0

1 Like
6

Great stuff! As always :wink:

I was wondering if you could add a list of things community would be able to do with this LLM

7

I believe that the major advantage for the community is having access to an LLM where your data is treated confidentially. But the list would be all kinds of querys done with gpt, for example. We’ll start focusing in text, though

All queries with AI go through mixnet, so the user’s IP and metadata are anonymized.

I also think that as one of the first sdk projects to come from the community, it will test the possibilities of integrating nym with other networks. I mean, what are the main difficulties and facilities in this.

2 Likes
8

I think this is a great proposal. What we’d need to assess better before proceeding, is that how self-sufficient can you be in the development → i.e. is everything up & running on the mixnet + SDK side that you need to develop this integration? How confident you are, that you can deliver without direct engineering support from Nym?

9

I’m confident in our ability to complete the project without direct assistance from the dev team.

Of course, we’ll be dialoguing with the Nym community and dev team through the available channels, but we’ve set aside a significant amount of time for testing and debugging precisely because of this scenario.

I think it can also serve as a feedback experience for the Nym team as to what the SDK’s greatest facilities and difficulties are and how it integrates with other ecosystems

:sunglasses:

1 Like
10

Great job @psydenst and squad, as always. This project is approved. Looking forward to seeing the first LLM implementation with user privacy built in!

2 Likes