Data as Labour Network (DALN)

All right, we're live. GM, GG, everybody. Welcome to what's going to be another great DevNTell. So if you didn't know, DevNTell is a 30-minute window for builders to showcase something they're passionate about or been working on in Web3. It could be an awesome project they've been working on, unit testing best practices, how to structure a project, smart contracts, automation goodies, etc. Basically, if you've got a passion for something, this is your opportunity to share it with the community. And today, I am super ecstatic to have Karen and Miffy on the show to talk about their project called DALN. Welcome to the show gang. Really appreciate you taking the time to come on here today.

Thank you, Narb. Yeah, GM everyone. Great to be back with, yeah, great to be back on DevNTell. Love to have you back. And this time you brought a new friend in Miffy. Yeah. I guess, before we get into the content, did you both want to give a quick introduction about yourselves in case folks aren't familiar? Sure. Yeah. So I'll start.

Hi everyone, my name is Karen. So my background is mainly in data science and ML. I've been working in the Web3 space since early 2021. And previously I have, I've led teams in building and launching projects in creator economy, NFT marketplace, and dev tooling platform. And I've been very active with the ETHGlobal and Developer DAO community. And with ETHGlobal, I've been a hacker, mentor, and judge. So I've played, I've worn lots of hats. And also with the Developer DAO community, I have had the opportunity to be a mentor in the past season. I'll hand it over to Miffy.

Hey everyone, I'm Miffy, and I'm a full stack software engineer with expertise in machine learning infrastructure and smart contracts. In DALN, I worked with Karen to develop a platform to consume the user data. And glad to see you. Excellent.

Yeah, well said both of you. And yeah, that's in just the time, let's get right into it. So I will bring your screen up, Karen. All right. All righty. I'll go for it. So we're really excited to present DALN. Are you able to see the slide properly? Yeah, we're able to see the slides. We see the notes assistant. Oh, okay. Yeah. Now you're good. Now you're good. Yeah. Okay. Great.

Yeah. Okay. So we're here to present DALN, Data As Labor Network. So it's a data DAO for collective ownership and monetization of consumer data. So first of all, I'd like to give a shout out to my amazing teammates. And we have definitely received lots of love and support from the Developer DAO community. A couple of our contributors, Mark and Kevin, are both Developer DAO members. And besides them, we have Miffy here, and there's Yifeng, who is a very talented designer.

So first of all, I want to share what inspired us to work on this project and where the name DALN comes from. So it stands for Data As Labor Network. So there's this book that I highly recommend and love, 'Radical Markets'. So basically there's been a reckoning and this paradigm shift that, well, in the past two decades, we as consumers and users have been accustomed to this business model that software products are offered to us as free products. But on the other hand, our digital print and all the data we generated on the internet are being used as free capital by Web2 platforms. So with DALN, with data as labor concept, there's this paradigm shift that, well, we are perceiving data as a labor that should be fairly compensated instead of free capital that could be used by Web2 applications or ecosystem.

So what exactly is consumer data? So consumer data is a fairly generic term, apparently. So it can be any data that applications or platforms collect about a user, about their preferences, behaviors, and interest. So on the individual level, those kind of metrics and data power things like personalization, personalized ads, or recommender systems. And as we know that those are really trillion-dollar businesses. But when we combine those data together, they could be even more valuable and can be really insightful leading indicators for the economy. For instance, if we look at credit card transaction data alone, if we have an aggregation of 500,000 credit card users' data, and we're able to easily find out, hey, is there a, any trend in the spending patterns, behaviors at hardware stores? Or hey, is there a leading indicator of Netflix subscriptions right before their earning announcement? So this powers the so-called alternative data market. It's a market that's an industry that's been growing really fast. So the valuation of the entire global alternative data market is estimated to reach almost $144 billion by 2030, and it's expected to grow at a 54% CAGR. So among all these different categories of alternative data, credit card transaction data is of particular interest to myself and to my team. One thing is that it accounts for arguably one of the most lucrative categories in the entire alternative data market. And also there's some interesting stats that I'd like to share about credit card transaction data. So worldwide, almost 1.86 billion credit card transactions are completed on a daily basis. And within the US alone, 184 million American adults, so that's roughly about 70% of the US adult population, have at least one credit card. And within the US, about 70% of retail sales nationwide are actually being done with credit card. So this is why we are particularly interested in how credit card transaction data is being traded and monetized in general.

So even though despite the fact that our personal data, particularly our credit card transaction history, is extremely valuable when it's combined together, we face a conundrum. First of all is that we as individual customers or consumers, there's no way for us to find out where, how, and where, how our data is being used. And another side of the coin is that if we're able to sell, if there were such a platform or marketplace where we can list our data, the paradox is that the marginal value of one person's data is close to zero. So this is why that lots of generic data marketplaces don't really work well in practice. And we're happy to discuss more about this in the Q&A section. So in both Web2 ecosystems, there are already a few projects or applications that explore this possibility to monetize credit card transaction data in aggregate. And those applications effectively serve as a data intermediary. But what our observation is that these applications or products, even though they're proclaiming that they want to give control and ownership back to consumers and reward them for sharing data voluntarily, but inherently they're centralized entities. And more importantly, how the aggregate data is being monetized is never transparent or traceable.

And this is why we advocate the concept of having a Data DAO. So now let's just zoom out a little bit and look at the concept and use case of Data DAO on a high level. So I think the audience of DevNTell are probably already very familiar with the concept of a DAO. Developer DAO is itself a DAO. So Data DAOs are basically DAOs that own and provision very large datasets. And those DAOs and as a DAO, the organization can have a governance charter and rules for curating and monetizing the data. And those DAOs could be either for profit or as a public goods. So the next few slides I borrowed from a presentation by Andrew, who is the CEO of Tableland, to illustrate some use cases of Data DAO.

So I think when people think about Data DAO, one typical, one common use case is DeSci. DeSci can be communities where they can crowdsource either research data or publications. And with the emergence of GenAI, we are now seeing lot more projects that come up at hackathons where they explore crowdsourcing data for either training or fine-tuning GenAI models.

So now let's get to the architecture of a Data DAO. So there are a few key components for Data DAO. So the first thing is a DAO membership that can be represented by a token. So this token could be just serve as only a governance token. It can be also serve the purpose of having certain utilities. And for sure, and also some other key components of Data DAO is data storage and indexing of the data as well as curation of the data. And of course, with the DAO, there should be governance structure in terms of how the data could be collected, curated, and monetized.

So now let's get to some more details about what we have been working on with DALN. So as I said previously, our main focus for this early stage is to build applications that have the capability for collective ownership and monetization of credit card transaction data. So we started the project in late last year. And for phase one of the product, our main focus was on tokenizing the credit card transactions and storage of the data on encryption and application of access control to data, as well as storage of the data on decentralized storage such as IPFS. And also there's an implementation of on-chain decryption that's verifiable on-chain as well as payment to data contributors. So those features are really good as a first step for a POC. However, we came to realize that for to order to launch the product for production and for potential mass adoption, we have to make some tradeoffs. For instance, the first one is that when the data when the encrypted transaction data is stored on IPFS, for sure, the data is immutable and tamper-proof. However, those kind of attributes also could be a hindrance to us because the data is static. But as a Data DAO, and in order to monetize the data, it's valuable to have evergreen and up-to-date data streams. So this is why when we design the phase two of our product, we decide to replace this decentralized storage of data contribution from individual DAO members with a cloud storage in a centralized database. And we can definitely share that in the demo in a little bit. And another thing we work on a lot, being working on a lot, is to add features that strengthen GDPR compliance. And definitely we can demo most of these features in the demo. So for instance, there's minimization of data collection and storage and usage of storage. And also through on-prem computation, we strengthen data sovereignty of data contributors. And also we have implemented we're implementing some solutions to have on-chain proof of data lineage. And also we have the rights to be forgotten is a very key principle in GDPR compliance. So for that, we have implemented designed and implemented some really interesting features that we're happy to share. And last but not least, we have incorporated ERC-6551 that help us to build token-bound accounts that makes the DAO membership a lot more interesting and have so-called configurable utilities.

So let me just gloss over this architecture design and I think with the further ado, I will just jump to the demo. Narb, are you able to see this page, the landing page of the DApp? Yep, yep. Okay, great.

So we have deployed the app on calibration testnet of FVM. So it's Filecoin's own EVM compatible chain. So just disclaimer, the FVM is relatively slow than other L2 chains. So just for this demo, please bear with us because some transactions may take little bit of while. So here we come to the landing page and now we have a wallet on the testnet of FVM. So now let's go ahead and collect a wallet. Okay. So this is the page when a user consents to the TOU and privacy policy, he will be prompted to join the DAO.

And the first step of this onboarding process is to for the user to authenticate their access to their own credit card accounts through Plaid API connection. So here we implemented Plaid in a sandbox environment, so all the data here is mock data and we're not using any real credentials of credit card accounts. So here I'm just going to put some mock-up information in order to pull the data. So just know that well here, the we are not capturing so the the users are not required to share their actual card number or account information. So we're not capturing any of those such information. Consent. So with this mock data, we actually check the checking account in the in the pop-up screen, but in practice, the user will only need to share their to authenticate and access their credit card account.

So now this authentication process is already has already completed. We move forward. Now we come back to the onboarding page. It's going to take a little bit. Okay, now we see this progress bar. So here, this is still work in progress. So the screen you see here is still a mock-up. So what's going to happen in under the hood when it's done is that well the the JSON file we just pull we actually just pull with the authentication of Plaid API for this given user's account. We're actually pulling a JSON file from Plaid. So here on the screen, I'm showing this mock data which just pulled from the API. So you can see that the data is ready it's already anonymized and there's no PII, there's no personal identify information here. And for every transaction, there's information such as the transaction date, the the geocode of the the merchant if if it's available, and categories of of this purchase, and and like some other information. So what's under the hood here is that well this JSON file is now being downloaded to to the browser cache and we actually run a computation node remotely and we run the analysis of the data on-device on the device of the user to to aggregate data. For instance for here, what the information we can collect from the user is that we're not collecting the full transaction data here from the user. Instead, we aggregate data by quarter or by category just to get a sense of the the behavior analysis of the given user. So once the analysis is done, the user will be prompted to sign a hash transaction to prove the provenance of this data stream and move forward and then this data analysis will be uploaded to our centralized cloud storage. After this step, the user will be prompted to mint a DAO membership token to officially join the DAO. So this token by design is a non-transferable token, aka soul-bound token. Now let's confirm this transaction.

So is the idea while we wait for the transaction to complete, is the idea that this is like a one-time operation? So when I sign up for DALN, I only need to upload or give consent to upload my data once and that'll port in all the transaction up to today, but what about the transactions that happen in the future? Do those get ported in?

Good question. So what we capture from the user is a Plaid token, it's a token that authenticates that give us access to the endpoints of their own credit card accounts through Plaid. But there's something that we also want to design a feature that that gives user a UI for them to update their data contribution periodically. I think Miffy can chime in on how this can be done on the engineering side.

Yeah, so basically it's we store a user's Plaid token ID in our database, and every time we can directly query this token ID to update the data. Gotcha.

Okay, great. Yeah. So now we had already minted the soul-bound token. The next step is to convert this token to a soul-bound token we just minted is essentially a ERC-721 token, right? And then the next step is to convert this token to a token-bound account through ERC-6551. So there's some delay in this step, so just bear with us for a second. There's some bug in triggering this action. It needs some time. Okay. Actually, we're going to show you something while we wait. So we have also incorporated Tableland, which serves as a decentralized SQL database that stores and captures the the change of the DAO membership. So here we have just minted the soul-bound token but we do not know yet the token ID of the token we just minted, but later on when we move on to the dashboard page we'll be able to see. Okay, now we're doing it. Okay, we just trigger this conversion to token-bound account.

So up to this point, the user has consented to give their data from Plaid and now after this step, they're a DAO member because they hold the membership with the token, correct? Correct. And then I guess at this point, perhaps this is coming while we wait for the transaction to complete, but how how would a person get rewarded for giving consent to DALN to have their data at this point? Or does it kind of go on a marketplace or something like that?

So there will be so we have on our product roadmap, we plan to build a DAO governed data marketplace. But that data marketplace is different in the sense that it's going to be a so-called on-demand data marketplace. Basically only someone if when the volume of members when the size of the DAO reaches a certain critical mass, for instance when we reach for instance 100,000 members in the network, so the aggregate data is valuable enough to be listed on the DAO governed data marketplace. So interested buyers could come in and negotiate or settle a final price with the DAO, either through auction or fixed price. And what we have envisioned is that there's going to be escrow mechanism where a deposit will be where a portion of the the agreed-upon price will be posted into a escrow smart contract. So this is going to trigger the DALN as a intermediary in the processor to go ahead and gather all the the full transaction data from all the users to get the up-to-date data from everyone and then do the processing aggregation and then leverage things like encryption and IPFS storage to provide that data to the buyer. Interesting. Cool.

Okay, now this is done. So now let's go ahead to the dashboard. So this dashboard by design is token-gated. So after someone joins the DAO, when they come back to DALN, instead they will be once their token is verified in their wallet, they'll be routed to this page whenever they come back to the application. So here I want to show you this. So this is the the contract address of the token-bound account. So basically this is a wallet that's owned by this that's tied to this particular token that this member owns. So in the future, what we envision is that we're through some partnerships, we can airdrop either NFTs or ERC-20 tokens to this token-bound account to reward users. So with this account, so and because this is a contract address, this is a wallet address, the user can always see the content of in this wallet on OpenSea and this could be cross-chain.

For instance here, this soul-bound token, the membership token is now deployed on FVM, but on the other hand, the airdropped tokens or NFTs could be on other chains. So one last thing to demo is the process to burn the token. So this is going to happen if someone want to exit the DAO and stop sharing their data. So they can just initiate this token burn event. And because this is an on-chain interaction, they can always verify this in a explorer. And again, this step is also the token burn event is also synchronized with the Tableland database. It's 18, so because the token is not burned yet, we can still see the entry for token ID 18 in the Tableland database. Okay, so now this is done. It may take a bit while but I think in a few minutes if we refresh the Tableland database we can see that well this entry for 18 should be erased from the database. I think yeah, that's it for the demo. Yeah, we'll open up the floor for Q&A.

Yeah, that was awesome. I guess yeah, there's a lot of potential here because I mean data is king, Web2 or Web3. In Web3 it's just going to get be even more prevalent I think. So when I guess maybe you two are still thinking about this, but when you do eventually launch the marketplace and a person sells or wants to sell their data to or just wants to sell their data, can only one person buy that data set or can multiple entities purchase that data set at once?

You mean on the purchaser side, on the buyer side? For sure, because basically it's going to be a data licensing model. The same set of data could be licensed to multiple entities. There's no exclusivity there. Yeah, totally. Yeah, makes sense. But there may be variation in the price in terms of like the recency or the length of period covered in each data. Gotcha.

And is the plan to always stick with credit card data or are there any plans to kind of expand the types of data that you would want to market? Yeah, great question. So I think we definitely we want to expand to other other types of data in the in the future, but for this initial for this early stage, for this beta, we want to mainly focus on credit card transaction because it's fairly easy to get, it's highly standardized through Plaid API, and also because it's as a data source, it has very high signal-to-noise ratio. So it's fairly clean and give us lots of information about someone's behavior, behavior patterns or preferences. Yeah, 100%. I mean we basically are configured to be transaction machines whether we want to or not.

I guess before we conclude, what what was the biggest challenge you and Miffy kind of went through building the first version and second version of DALN? Can you speak to that? So I think for the for the V1, the V1 product, the challenge the biggest challenge was to was the implementation of decryption encryption and decryption process. And for the second one, for the phase two product that we're working on, it's I think one it's really on the conceptual level is how we make tradeoffs for certain things, like as I illustrated in the deck, right? How much decentralization we want to get want to achieve through the product as opposed to how this product is optimized for for future usage.

Gotcha. And now for sure and also for this for our planned beta launch, another thing is definitely friction for onboarding users. Because we're excited about this product exactly because it appeals to it should appeal to non-Web3 native users because as I said within the US alone, 70% of adults own a credit card, so they can all be potentially the customer user for this application and service, right? But this is something that all consumer application in Web3 face, how we can eliminate frictions for onboarding users. Yeah, 100%. It's always the challenge, but I think eventually we'll end up overcoming. Yeah, like a low-hanging fruit is the first one is low-hanging fruit it's just to have to enable gasless transactions so that well even though the membership the DAO membership token is free to mint, but someone doesn't need to deal with the hassle of buying some crypto token just to cover the gas fee to get the token, right? Exactly. Yeah, 100%. I agree with you. Yeah, that's the biggest thing is just like having to ramp yourself up with like the wallets and the tokens and transaction signing just to start using Web3 is kind of a pain so the more we can abstract that away from the user the better and the better chance we have to like attract everybody, not just like Web3 folks right? Yeah. Totally get you.

I guess before we conclude, what's the best way for people to keep track of your progress, get in touch with you if they're interested, asking you questions and what not? Sure. So let me see on the last page of deck, we have two QR codes. One is our Twitter, one is a Telegram group for closed beta. Awesome. And yeah, we'll be sure to. Yeah, so feel free to DM us on Twitter. Yeah, for sure. Yep. And the this information will be on the description of the YouTube video as well. So if you don't have a chance or didn't have a chance to scan it, it'll be there as well.

But with that, just want to thank you both, Miffy and Karen so much for coming on. Really, really appreciate you both giving us a chance to get a sneak peek at DALN and really looking forward to see how the product evolves and you're always welcome back on the show to show the next phase of DALN. Well thank you for having us. Yeah. My pleasure. And before everybody leaves, I just want to share a QR code of my own. Bang. So for folks who stayed and watched to the end, this is your QR code to scan and claim your NFT for being a viewer of DevNTell today. So what you'll want to do is get your regular QR code scanner, scan this QR code, fill out the form, and you will be airdropped a DevNTell viewer NFT within the next couple of days. And you'll have an hour and a half from this point to be able to claim that. So be sure to get on that if you want to get your NFT for being a viewer today. And with that, just want to wish everybody a very happy Friday, if it's still Friday for you, happy weekend, and that will catch you all back here next week. All right all. Have a good one. Great. Bye everyone. Bye.