Algovera: A Community of Data Scientists building Data Science Apps

OceanDAO Round 12
#1

GAN-training
GAN-training1367Ă—454 973 KB

Part 1 - Proposal Submission

Name of Project: Algovera

Proposal in one sentence: Making it easier for data scientists to build data science apps on Ocean by integrating with existing framewokrs, running weekly hacking sessions for app building and inviting grant proposals for data science projects on Ocean

Description of the project and what problem is it solving:

Currently, there is a lack of data scientists in the Ocean eco-system and Web3 in general. Algovera is a data science DAO onboarding data scientists to Web3 and training them to use tools such as the Ocean marketplace, Compute-to-Data (C2D), and more through our weekly hacking sessions and other community resources. Through our weekly data science hacking sessions, we have been running a Proof of Concept (PoC) for a data science team to build and publish an algorithm, and share ownership through a DAO (that governs the flow of value generated by the algorithm). Now we want to make it easier for data science teams to create data science apps (on top of the algorithms that they develop) by integrating with existing frameworks such as HuggingFace and running weekly hacking sessions for app building. We will upload the CryptoPunks dataset and our generative art algorithm to HuggingFace under the Algovera organisation and explore the possibilities of connecting the datasets and algorithms with Ocean marketplace in the backend. We will also create a PoC app with a backend that serves the algorithm from the Ocean marketplace. We will also use $5k of the funds for grants for data science projects on Ocean (voted for by the Algovera community), which we will advertise widely in our community, HuggingFace, fast.ai, EleutherAI, Kaggle and other data science communities.

Our growth over the last month or so:
Discord 54 => 113 members
YouTube 6 => 20 hackers
YouTube 13 => 24 maximum people at a single session
GitHub 3 => 6 contributors
Twitter 45 => 108 followers
Twitter 2.1k => 12.5k impressions

Grant Deliverables:

  • [ ] Upload CryptoPunks dataset to HuggingFace (see example dataset)
  • [ ] Upload DCGAN generative art algorithm to HuggingFace (see example algorithm)
  • [ ] Create 1X app with backend that serves recently-published generative art algorithm from the Ocean marketplace (e.g. using Gradio)
  • [ ] Outreach with the HuggingFace community
  • [ ] Publish blog post on creating web apps with Ocean marketplace in the backend
  • [ ] Run weekly hacking sessions for building a data science app
  • [ ] Continue running weekly data science DeFi hacking sessions
  • [ ] Publish Discourse forum for long form discussions and proposals
  • [ ] Implement SourceCred for tracking objective contributions (on GitHub, Discord and Discourse) to data science projects
  • [ ] Implement Coordinape for tracking subjective contributions to data science projects and use to pay one contributor
  • [ ] Set up Gitcoin bounties (for contributions like blog posts)
  • [âś“] Publish community directory of active members
  • [ ] Finalise data science grants program for Algovera community
  • [ ] Host at least one vote on grant proposals and fund at least one data science project
  • [ ] Update website (e.g. with open positions, blog posts) and advertise engineering role

Which category best describes your project?

Build / improve applications or integrations to Ocean

Which Fundamental Metric best describes your project?

Data Consume Volume

What is the final product?:

A web app for our generative art algorithm that uses the Ocean marketplace in the backend. Using the web app generates consumption of the algorithm on the Ocean marketplace, which directs value back to the members of the project DAO.

We are planning to use the Gradio framework to maintain similarity with the popular HuggingFace Spaces. In future rounds, we hope to create a similar data science app marketplace on top of Ocean. Further into the future, we envision something like scaffold.eth for data science apps.

Value Add:

  1. Usage of Ocean

Creating a web app for our algorithm will increase consume volume of our algorithm data token. Furthermore, it will act as a PoC for more streamlined creation of other data science apps. Using the Gradio framework will hopefully help us with outreach to the HuggingFace community. We will invite proposals and advertise our Algovera grants programs for data science projects with $5k of the current grant. This will result in one or more new projects and assets using Ocean.

Funding Requested: $28k USD

Proposal Wallet Address: 0x81e80B838e04644a93e4F3C701038889d3a1b95e

Have you previously received an OceanDAO Grant (Y/N)? Y

Team Website: www.algovera.ai

Twitter Handle: @richardblythman

Discord Handle: richardblythman | Algovera.ai#3425

Project lead Contact Email: richardblythman@gmail.com

Country of Residence: Ireland

Part 2 - Team

Core Team

All three of us are working full time on this project.

Richard Blythman

Hithesh Shaji

Keaton Kirkpatrick

Part 3 - Proposal Details

Project Deliverables - Category:

IF: Build / improve applications or integration to Ocean, then:

Code for web app with Ocean backend will be made available at https://github.com/AlgoveraAI.

If the project includes community engagement:

  • Running the campaign on social media for how many weeks?

We will continue to run several weekly sessions: (i) data science app hacking sessions, (ii) data science DeFi hacking sessions. These sessions are recorded, uploaded to our YouTube channel and publicised on Twitter and LinkedIn.

Project Deliverables - Roadmap

  • Any prior work completed thus far?

Last round deliverables:

  • [âś“] Publish 1X generative art algorithm on the marketplace and distribute tokens amongst contributors
  • [âś“] Start a DAO to govern development of the algorithm, distribution of tokens etc.
  • [âś“] Documentation on stack of DAO tooling to be used to coordinate and govern algorithm development in a decentralized working group
  • [ ] Host 1X vote on Snapshot among algorithm token holders
  • [âś“] Interview 2X artists about bringing their existing generative models or datasets on-chain
  • [âś“] Continue running weekly data science hacks (where we go through the whole process of training a ML model using new Web3 tools, publish it and share ownership). Recordings here.
  • [âś“] Continue running weekly community calls introducing data scientists to the idea of decentralized AI . Recordings here.
  • [âś“] Publish blog post on bringing generative art models on-chain with Ocean
  • [âś“] New Website (full rebuild) here.
  • [âś“] New Discord (full rebuild) here.
  • [âś“] New Notion (full rebuild) here.
  • [âś“] Pitch deck here .

Hacking sessions for ML algorithm development: The workflow and practices for data scientists are quite different using new technologies such as marketplaces and private AI, and feedback has indicated that data scientists don’t know how to dive in. We have now held 7 data science hacking sessions that have been attended by 20 unique individuals. 6 of these focused on our PoC for a generative art model where the aim was to go through the whole process of developing an ML algorithm using the Ocean marketplace and C2D. We also just held the first hacking session for data science on DeFi data series last week. You can watch all of the recordings on our YouTube channel here.

Generative art algorithm published on the marketplace: We published our generative art algorithm on the Ocean marketplace (here). We have tested that this algorithm runs by training it on the CryptoPunks image dataset (here) using C2D. You can find the Python code for the algorithm on our GitHub. There are also an accompanying set of notebooks. The repo itself is well maintained with a Kanban board for contributors. Our community members have uploaded demos of other models including models to convert images to artistic sketches and to convert text to images (using VQGAN + CLIP).

DAO launch for project contributors: It is important to us that the developed algorithm is owned by a community of individuals rather than one centralized party. Furthermore, we want data scientists to gain experience with Web3 processes such as coordination and governance. Just using Web3 tools does a lot of the heavy lifting. With the above idea newcomers would create and use digital wallets, use a shared treasury, take part in voting, and more. We have created documentation on the stack of DAO tooling we are using to coordinate and govern algorithm development in our decentralized working group (available here). We launched an Aragon DAO and are distributing the tokens equally among the 21+ contributors.

Generating consumption of generative art algorithm: We are working on generating consumption of our generative art algorithm. We’ve been interviewing artists, creators and art DAOs to explore interest in our PoC that unlocks more value by bringing access control for generative art algorithms on-chain. One of our leads involves acting as ML Web3 consultants for a simple PFP project for an art DAO. We are also exploring the requirements of artists and creators for bringing their existing datasets and algorithms onto Ocean.

Blog posts: We have written a blog post about our innovation for unlocking more value for NFT enthusiasts using Ocean (available here). We are also posting a bounty to the community to write a blog post on our series of hacking sessions for generative art.

Community discussions: We’ve held six community discussions (weekly on Tuesdays) on the decentralization of AI, AI governance, AI ethics and AI for tackling climate change. You can watch most of the recordings here. Trent attended the most recent session where we talked about all things Decentralized AI, AI DAOs, AI in Art/Climate/DeFi, Tokenizing AI communities and more.

  • What is the project roadmap? That is: what are key milestones, and the target date for each milestone.

Dec 1: First hacking session for data science with DeFi data
Dec 7: Publish community directory
Dec 8: Hacking session for data science with DeFi data
Dec 9: First hacking session for building a data science app
Dec 13: Publish Discourse
Dec 14: Post Gitcoin bounty
Dec 15: Launch data science grants program and invite proposals
Dec 15: Hacking session for data science with DeFi data
Dec 16: Hacking session for building a data science app
Dec 20: Host vote on data science grant proposals
Dec 21: Finish implementing SourceCred
Dec 22: Hacking session for data science with DeFi data
Dec 23: Hacking session for building a data science app
Dec 24: Publish updated website and advertise new role
Jan 3: Finish using Coordinape to pay one contributor
Jan 3: Announce results of data science grant voting
Jan 6: Publish web app
Jan 6: Publish blog on building data science apps with Ocean backend

  • Please include the milestone: publish an article/tutorial explaining your project as part of the grant (eg medium, etc).

We will publish one blog post (targeted at data scientists) on building data science apps with Ocean backend.

  • Please include the team’s future plans and intentions.
    • Any maintenance?
    • Foreseen or possible additions?

We hope to develop algorithms and apps in different domains including face anonymisation and motion analysis.

We are working towards creating a web app for data scientists that includes a data, algorithm and data science app marketplaces (similar to the popular HuggingFace) with Ocean in the backend. We hope to include some of the cool functionality of the HuggingFace dataset and model marketplaces such as previewing and exploring data, and also the ease of creating data science apps on HuggingFace Spaces. We also want to explore creating functionality to use Ocean with the HuggingFace app. Imagine an option to train an algorithm using Ocean C2D on HuggingFace:

hugging
hugging1598Ă—394 67.4 KB

Eventually, we envision something like scaffold.eth for data science apps.

#2

Its nice with the initiative on HuggingFace. We also think this is the right front-end to bring data scientists in, consuming data easier, rather than dealing with how market mechanism works.

1 Like
#3

I my view, I think that Ocean needs to focus more on two things that are severely lacking:

  1. more people who are actually doing real data science and not just creating apps for data markets or writing reports.

  2. actual high quality data sets that are well curated and comprehensive

In terms of algovera - they are a genuine team of people with real credentials in actual data science and Machine Learning. I do believe they have the capacity to grow a meaningful working party to explore real data science and ML challenges and they can be critical for the Ocean ecosystem to facilitate the real users of Ocean if it is going to succeed. In my view failing to vote for this team is a mistake if you are an ocean holder - these type of teams will actually bring real value and use.

3 Likes
#4

Hi there,

For transparency, starting R12, all proposals will have to be funded within 2 weeks of winning a grant.

The funding deadline is December 27th 23:59 GMT.

You can read our wiki and how to submit a funding request to learn more.

1 Like
#5

Thanks Amir. We’re hoping to interface with many data science frameworks in the long run and HuggingFace is one of the biggest. It’s still too hard to do data science on Ocean.

Happy to keep you synced in or collaborate on HuggingFace initiatives if you’re interested over at Datalatte :slight_smile:

3 Likes
#6

Thank you Gareth for the very generous vote of support. It means a lot from someone with your vast experience. Genuinely appreciate it.

We agree that increasing the data science and ML expertise in the Ocean ecosystem is critical to Ocean’s success.

2 Likes
#7

Hi everyone,

It was brought to my attention that this team has completed 3 Grants enabling it to access a maximum of 20k per funding round. However, for R12 it has required 28,000 OCEAN or 8,000 OCEAN above the limit.

It seems no one picked up on this however I am reaching out to let you know in case your request has not yet been made, and request the maximum amount instead. In the case the invoice has been submitted, I would propose the team continues to work with this in mind and makes the appropriate adjustments in consideration of future funding.

We will be working towards improving the grant UX & engagement from the Project-Guiding WG to help facilitate the grant process and make sure all projects are operating in accordance to the guidelines. I agree that this information is not easily accessible, and am making recommendations to improve it.

All the best.

#8

Ah, I didn’t notice that the jump from Experienced Project to Veteran Project was 2 completed projects to 5 completed projects. I thought it was 3 completed projects, but I now see this changed after Round 10.

I had been discussing with @AlexN last round how to manage our fork from Data Onshore, with whom we completed 3 grants with me as co-lead and Hithesh as a developer. We decided it didn’t matter since VisioTherapy had also completed 2. But actually it does make a difference for us based on the 2 => 5 requirement. Algovera is a direct follow on from Data Onshore. It quite literally builds on the code we developed. Is there some way that Ocean can take into account forks like this, or does the forking team go back to zero?

#9

Brief / Problem

Funding tiers right now are calculated based on projects and their ability to complete grants.

This is currently something that needs to be done by hand, impacts the grant history, and needs to be verifiable. If a good UX was in place, you might have been communicated that your project might have been rejected as dependencies were not addressed in the DB.

Solution

I think the solution here would have been to roll (Data Onshore + VisioTherapy) into [Algovera] so all previous grants from those 2 projects would have registered as being completed & counted towards Algovera.

@richardblythman, could you perhaps post about this in the Discord #PGWG so we can address it there? I don’t think this is an issue as it’s the same team, I just want to involve others in the process so we can discuss this more easily, have some consensus, and get this sorted out.

Looking ahead

There is currently no feature to “Merge Projects”, aggregate their history, and maintain a data provenance to provide this retroactive insight. In the future, this will be handled via a contract, so we should look at this use-case as a way to make sure that teams are able to rebrand/restructure/leverage past successes.

#10

OK great. Yes, sure. Will do that. Thanks :slight_smile:

1 Like
#11

[Deliverable Checklist]

  • [X] Upload CryptoPunks dataset to HuggingFace
  • [X] Upload DCGAN generative art algorithm to HuggingFace
  • [X] Create 1X app with backend that serves recently-published generative art algorithm from the Ocean marketplace (e.g. using Gradio)
  • [X] Outreach with the HuggingFace community who also use Gradio
  • [X] Publish blog post on creating web apps with Ocean marketplace in the backend
  • [X] Run weekly hacking sessions for building a data science app
  • [X] Continue running weekly data science DeFi hacking sessions
  • [X] Publish Discourse forum for long form discussions and proposals
  • [X] Implement SourceCred for tracking objective contributions (on GitHub, Discord and Discourse) to data science projects
  • [X] Implement Coordinape for tracking subjective contributions to data science projects and use to pay one contributor
  • [X] Set up Gitcoin bounties (for contributions like blog posts)
  • [X] Publish community directory of active members
  • [X] Set up data science grants program for Algovera community
  • [X] Host at least one vote on grant proposals and fund at least one data science project
  • [X] Update website and advertise engineering role

Thanks @idiom-bytes.

1 Like
#12

Hi @richardblythman ,

Thank you for submitting an update for your previous proposal!

As a Project-Guiding Evaluator I have reviewed your Grant Deliverables, and they look to be in good condition.

I have registered your latest updates into Airtable and also looked at your Project Standing, which looks to be in good condition. You are clear to apply for another grant.

I would like to thank you for your positive contributions to the Ocean Ecosystem and I look forward to reviewing future proposals from your project.

All the best!
-Idiom

1 Like