ChainBrain

All right, we're live. GM, GM 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 are passionate about or have been working on in web3. This could be an awesome project you've been working on, demonstrating testing best practices, automation goodies, smart contracts, how to structure a project, etc. Basically if you've got a passion for something, this is your opportunity to share with the community. And today I'm ecstatic to have Marissa on. Welcome Marissa.

Thanks for having me, I'm excited to be here.

My pleasure, my pleasure. I'd love to have you on. Marissa's going to be talking about ChainBrain today. So but before we get into that, did you want to give a brief introduction about yourself, how you got into web3, what you do outside of web3, and how you found yourself in Developer DAO?

Yes. Um, I originally come from a data science background. I was working as a data scientist at a fintech startup building machine learning models on transaction data. Um, and on my on the side I was trading, um, and was interested in understanding the data behind tokens, NFTs, etc. So from my own curiosity and as someone with the data background, I went on Dune and other analytics platforms to try to understand the data. And through that, I understood that web3 data was way more complicated than it seems. It's really hard to know where you're actually supposed to pull the data from or what it represents. And that's kind of where I had the idea for ChainBrain, um, to build a um, real-time analytics, predictions, and on-chain actions platform. So ChainBrain's a business intelligence tool that users can search and visualize blockchain data. So I can present a little bit more about ChainBrain and then I'm going to show you guys a demo.

Excellent, love it.

Cool. You guys should be able to see my screen. Yep. Um, so like I mentioned, ChainBrain is a web3 business intelligence tool for real-time analytics, predictions, and we're even building out something really cool for executable on-chain actions. We came out of ETHDenver in February, um, where we won a couple bounties including Polygon NFT analytics and The Graph's bounties. So a bit about the team, so my background is as a data scientist, um, and we have Prakash who is a frontend engineer and Hope who's an ML engineer. So kind of the problem that we're seeing is that companies spend a lot of money on manual data aggregation from many sources. So for example, if I'm a DeFi protocol and I need to not only pull in data from Dune, Etherscan or another block explorer, but also maybe I want to historically price that data in USD. So now I need to either hit a pricing oracle or for example CoinGecko's API and price that transaction at the time that it occurred in a standardized currency whether that's USD, Euros, etc., to understand um, kind of how my um, value is moving over time. So companies spend a lot of money on setting up this data pipeline. Um, it takes engineers a long time to not only set up all the different data sources that they want to track, um, but manage the smart contract addresses. Some companies even track over 250 contract addresses, um, depending on the complexity of what they've built. And all of that comes with a lot of manual maintenance.

So that's where ChainBrain comes in and helps um, developers like um, users understand their on-chain data. So it's really um, ChainBrain is a business intelligence tool that allows you to make predictions and analyze on-chain data, um, as well as take action off of that data. And that's something that we haven't built out yet, but we are working on. So one of the things that I've really thought about is so now I have all this data in a nice dashboard and what do I do with it? What actions can I take off of it? So I could be a web3 social company and let's say I have like the top hundred posts, for example, and let's say I want to add them to a leaderboard. I can do that automatically through what we're building with ChainBrain. Um, so I can execute an action on-chain based on that. Same with if I'm a DeFi protocol and let's say I have a liquidity pool and every time the liquidity goes under a million dollars, for example, I want to add more liquidity. I can do that automatically through ChainBrain. So we're not only monitoring and providing alerts about on-chain data, but we allow you to take action on-chain based off of that data. So that part is not built out yet, but I'm going to show you what we have built out.

Um, can you you guys can see my screen now which shows ChainBrain, right? Yes, yep. Okay awesome. So let's say that I'm interested in lending and I want to search um, the Uniswap subgraph. So currently we're go on top of The Graph and um, that's just one of the ways that data is retrieved from the blockchain by using like The Graph protocol's subgraphs. So how it works is I would select like the subgraph that I'm interested in, Uniswap, and I can type in my query. At the moment we're displaying the GraphQL um, to testing um, we're going to have that be able to be toggled off, but we're testing how do users like that, which type of users are using our product. The goal is for end users to be like business-level users so we would abstract this away, but a lot of our beta testers right now are developers and with that we want to provide the actual code behind it so that they could go into like a GraphQL editor and play around with it. Um, so I select my Uniswap subgraph and let's say I'm interested in finding the top five USDC pools by volume. I want to get rid of my USDC through Uniswap and I can look at the top five pools by volume and see which one I like the most to exit from my position. So you could I could just use the biggest pool which has the least slippage, but in this case, um, the biggest pool is like ETH which has a lot of volatility. So maybe perhaps I want to do a different pool um, which has like maybe the the third one, um, which has a very low fee. Um, so I can use this to help understand and um, then I could go onto Uniswap and choose my pool, for example.

So another example is let's say that I am an NFT trader and I want to look at the daily volume on OpenSea. So I can then select the OpenSea subgraph and I can type in um, show oop sorry, show the volume um, of all trades of all trades by day in um, yeah by day. Show the volume of all trades by day. Let's see. So I'm I'm interested in understanding um, the amount of items and collections traded that day. And so what this does in the backend since you guys are all a bunch of developers I figured I can um, tell you, um, is we use GPT to take the natural language, generate it, generate GraphQL and then we pull this data from the graph. So basically abstracts away the coding from the user.

So now I can see that all of these um, these volumes by day are recorded and I have the timestamps, the collection counts, the item counts and the cumulative trade volume in ETH. Um, so I have all of these different metrics. Um, and let me think. So we could also do a governance use case. So if I look at Compound Governance for back in March of 2022, Compound Finance had a proposal to add MATIC to the protocol. And this was proposal 86 and it was set to pass until the very end of voting, um, when at the last minute a16z swooped in and then struck down the proposal so it ultimately failed. So I could type in what was the voting timeline um, for against, for and abstain votes, um, for proposal 86?

So this will then show me the timeline, um, for that proposal. This was a um, an interesting case of when there's a lot of votes outstanding and you have a DAO you want to be able to like incentivize or alert your users, for example, of look there's an hour left in voting and 90% of votes are outstanding, come cast your vote. So then you can see like this is the moment when a16z came in and voted against proposal 86. Um, and I can save this to my dashboard and then go over to my dashboard and it will show me all this data. We're still working on the UI for this, um, but this is essentially what we have built out so far. Um, we are working on adding some different components. So one of the things that we're adding is time-based models. So GPT does not handle time like relative time well at all. Um, so we're building out an NER model for that as well as identifying like crypto-specific entities, uh creating like some mapping for that. Um, and one of the other big things that we are adding that currently we sit on top of the graph.

But one of the things that we want to add in the future is uh the ability to aggregate your data from any source because that's one of the issues with a lot of the um dashboard or visualization tools today is that you have to use the data that they provide. You can't choose your own data. And in web3 which is, you know, we're all about decentralization and giving the control back to the user, I want to choose where I want my data to come from. Maybe I don't like Dune's data, maybe I don't like Etherscan's format of their data. And maybe I want to connect directly to an RPC provider and then visualize on top of that. So what I would have to do currently is build out that entire pipeline myself and then pipe that data into what is traditionally a web2 tool such as um Google Data Studio, Looker, etc., um which many web3 companies are using and visualize it that way. So we're basically having to create that entire pipeline and then stick it in a web2 tool.

So that's kind of what we're trying to build here at ChainBrain. Um, and so this is a little demo video, um but I showed you the live demo so we don't need to to go through that. But I really want to talk about kind of why this is important. I think that businesses spend a lot of time manually creating these data pipelines, um both time and money in terms of like salaries, um and web3 businesses in order to compete and to grow, they really need to understand their data, um and be able to take actions off of that data. One of the really cool things about on-chain data being so transparent is that you can not only see your data but you can see your competitor's data. So I could I could look at like TVL of a bunch of different DeFi protocols, for example. Um, so we really believe that now is the time for this, um that blockchain data should become less complex and allow more users to analyze it.

So I know there's a bunch of other BI tools out there, so I just wanted to kind of go over that and position ourselves in that market. So ChainBrain is an operational product that is designed to be used by business users, not consumers. We are really focused on helping web3 businesses succeed. Um, and we also really believe in openness, openness of data and modularity. So in the current state of web3 analytics products, you have to rely on the the product itself to parse the data you need or you have to parse it yourself like I mentioned and then put it in like a web2 tool such as Looker. So we really want to change that with modular data source access. Um, so most analytics tools focus on some big analytics such as number of users or TVL, but what we ask what you can do with those numbers and we take analytics one step further with on-chain alerting. Um you can set up customizable alerts, executable on-chain actions and modular data source access because what good is analyzing your data if you're only analyzing part of it?

So our target customers are web3 startups that are interested in understanding their on-chain data, such as DeFi protocols, NFT projects, web3 social or gaming. Um really anyone that wants to understand their data, we can help them. So one our advantages like I've talked about is one we have these executable on-chain actions that allow users to um take action on-chain through analytics. And we're really focused on serving the whole data collection pipeline uh from from data collection to the actionable steps that you can take from that. Um so we really position ourselves as kind of a one-stop data shop for web3 data. Um and one of our future plans is to build out machine learning models on top of the data um to allow you to analyze and understand your data such as forecasting um churn, retention, all based on on-chain data insights and at at its core, um we want to help web3 businesses make better predictions so that they can stay in business longer. Um, so that was a bit about ChainBrain. I wanted to put out a little call to action out there. So we have um we aren't spreading our link around that widely, but I would like to extend it to the DevDAO community to test out our product. Um I will post the link in the chat, um now let me post the I can post the link. So this is a link that you guys can use um to test out search and query your data. Would love any feedback from that. Um I believe my yes my Telegram is on the screen @marissamega. Um we and secondly, we are looking for engineers well-versed in on-chain data to contribute to the project. There's a few ways that you can contribute, but some of them include building out functionality for um these executable on-chain actions, um data pipelining, data transformations. And I want to make a specific call out to if you have expertise in working with Solana data, um we have a grant for this work from Solana. Um so any engineers that are interested in getting involved, we would love to have you. Uh so please reach out to me on Telegram @marissamega. And thirdly, for businesses that need help with their data, we're looking for design partners who we can work with to develop added features that are tailored to your unique needs. So if your company is struggling to aggregate the necessary data and use data to make effective decisions, whether that's due to limited resources or other constraints and you would be interested in partnering with ChainBrain, please also reach out to me on Telegram. We would love to help you. Um we're in the initial phase of working with design partners and building out customized use cases for them. Um so we really want to work with you and we would love any help um. Our goal is really just to make on-chain data accessible to everyone. So if you're interested in that mission or passionate about that, please reach out, my DMs are open. Thank you. Um does anyone have any questions? I think I saw some stuff in the chat.

Yeah, we have a couple. I think you touched on one, but uh just wanted to say this looks awesome and it's amazing that this came out of a hackathon. I I can't wait to see the the fully polished product once you're once you've built out all those features you mentioned. Um but Simone had a great question, that do we also have templates to fill in the query line so that the user can be guided in the querying process?

That's a great um question. We are working on that. Um that will be a um an easy feature to add. Um so that's a great suggestion. We will we will add that. Um and Haribols asked are you guys hiring?

So we are currently yeah, we have um some grant funding and are raising a round, so we'll be hiring more uh in the near future. Uh but are currently looking for any contributors that want to get involved um within kind of the scope of some of the grants that we've gotten. Um so if anyone's interested please reach out.

Excellent. And uh are those uh open roles posted anywhere or should people just reach out to you with interest and then you can guide them?

Um you can reach out to me with interest.

Gotcha.

Yes. Um Haribols also asked are you guys building this full-time? Yes, um I am working on this full-time. Our ML engineer and frontend are part-time at the moment. Um so yeah that's the that's the current state, um but are also there's an opportunity for a full-time contributor um to come in and lead out some of the technical building.

Excellent. You heard it here first folks, uh get on that if you're interested in this type of thing. Um and if anybody has any other questions please post them um in the chat here and I will ask on your behalf, but uh till then I have a couple questions of my own. Um so the ML models that you use to create the um the GraphQL um are those specifically trained by you or can GPT do that kind of out of the box?

Yeah, we used a GPT endpoint for that. So we do we are using GPT. Um the one thing to note is that GPT we are using GPT plus some additional stuff that we've done. So for like relative time, we have like a specific NER model that will then like replace the time that GPT creates because GPT has no idea where we are in time like it was trained on data till 2021 and doesn't know what last week means. So for that we do some like replacement based on an NER model that we built out. Um so and same for some of the crypto-specific things. So it's GPT plus a mix of ML things that we built in-house.

Gotcha, gotcha. And I guess have you thought about uh I know you mentioned that you're specifically concentrating on web3 companies at the moment but have you thought about how uh ChainBrain could be used by uh web2 firms? Say web2 companies are starting to look at web3 and one big thing uh that stopping a lot of folks from getting in is the fact that to operate a lot on on these chains you need to effectively hold uh tokens and that sometimes doesn't play well with uh the the treasury folks right? Uh so is there is there a path forward um in your mind where uh ChainBrain can be used for treasury folks uh to be able to analyze how um token prices are going up or down and get suggestions on when to liquidate or when to hold or or whatnot? Can you provide examples of the predictions and actionable insights that ChainBrain generates?

Yeah, um so that's a good question. In terms of predictions, you could want to predict the amount of users that you will have three months from now based on your current activity and you could also analyze your competitor's activity to add as like a feature in that model for example, um or just compare them side-by-side. And actionable insights, so adding liquidity to a pool automatically based on it dropping by a certain amount or something like that, adding posts to a dashboard for a web3 social company automatically based on like a roundup. I could create um basically a function that gets um that updates my TVL across all of my contracts um and can then I can then have that in one place so I would be pulling data from multiple uh sources for that probably because I probably want to price that in USD. Um and another idea that you could do is if you want to um predict your um revenue or token sales you could look at some of the other things that your competitors have done in the past in order to understand how your token sale might go or your market making campaign for example. So they're just a few of the options. I'm very interested to hear if anyone has any other ideas for ways that you can uh create predictions and actionable insights on-chain or what would be helpful to your specific business. So if you have any of those, um definitely open to listening.

Great answer. Um and I guess in terms of um while you you three were building ChainBrain what were some of the biggest challenges that that you faced? Uh I'm sure uh during the hackathon there was more than one challenge. Um what's the process of integrating ChainBrain's business intelligence tool into an existing web3 infrastructure?

Yeah, um so we weren't so that wasn't the initial team that built it. Um we have an engineer, Gazudo and Keith and Prakash and I were all a part of the initial team that built it out. And in terms of building during the hackathon, I think we ran into a number of challenges. I think getting GPT to understand the queries uh themselves was tricky and we realized that we had to feed it a lot of examples um in advance for it to understand and be able to work with like non-examples in the future. Um so that's something we're still actually even working on that and trying to figure out how to create more example pairs of natural language input to GraphQL queries in a scalable and efficient way. Um so that was one. And then I think the other big challenge was the visualization component. We I tried a bunch of different automatic visualization libraries, autoviz, etc., um and none of them were really working. So we ended up writing something by hand that basically looks at the data types like if it's a time series it puts it in a chart, if it's XYZ it does a table, um and doing it that way but we didn't get to build out like the visualization side of things as much as we wanted to. Um so yeah those are a couple of the challenges that we faced during the hackathon.

Cool. Yeah, I mean you there's always some sort of friction whenever you're trying to build something innovative and cutting edge. So yeah it's great to see that the team overcame those. And I guess uh final question for me is uh your so ChainBrain obviously is is a mold of web3 and AI and we haven't really seen too much of that uh at least I haven't seen too much of that um coming forward uh but I think it's starting to kind of creep up now uh with the rise of GPT and how accessible um AI is now. Um what do you see kind of as um how things uh will play out in terms of AI and web3 kind of coming together as time moves on?

Yeah, I think that web3 and AI are kind of like the counterbalance to each other. Um I think that you need I think that web3 provides the accountability um the like logs the immutability that keeps AI accountable. Um and they work together in a lot of ways because one of the things about web3 is that it's very decentralized and transparent. And AI a lot of the times with big data, it's all this data is concentrated in the hands of a few companies, Facebook, Twitter, um Google, etc. They all have all this data and what web3 does is it decentralizes it and puts that power back into the hands of the individual. Um so I think they work together very well. And with like the rise in both of these technologies, um we are seeing I think more and more applications come out that combine both of them, um whether that's for applications using like ZKML, um supply chain security. There's a bunch of different applications that combine both web3 and AI. I think the future is very bright for both of them. Um I am looking forward to seeing more what comes up within that like domain, the crossover between web3 and AI. So should be um interesting. Embedding is on our roadmap. We have not um that's definitely further down our roadmap I would say, um to offer the ability to embed within your own custom platform. Like if I'm Lens protocol for example and I want to create like a dashboard that I want my users to have access to, we will eventually have embedding capabilities that will allow them to do that. I think for now like other business intelligence tools in the web2 space or web3 space that exist already, creating uh like it's as a standalone web app is is what we have at the moment. Um so it doesn't exactly integrate into the stack I would say, um or like whatever you're you're already using but I think people are quite used to going to a specific website or to a specific place to get their business analytics um yeah.

Excellent, yeah. And um unfortunately that uh brings us to time. Um but just want to thank you so much Marissa for coming on. This was a great spot. Uh love learning about ChainBrain and really looking forward to it continuing to be built out as um time progresses. And uh folks I'll share all that contact information with you in um in the accompanying data in the description of the YouTube video and the blog. Uh so you can reach out to Marissa if you have further questions or are interested in getting involved. Uh but in the meantime since you're all still here uh bang. Uh here is the QR code to claim your mint Kudos for being an attendee here today. So what you want to do is grab your QR code scanner, that's your regular QR code scanner, scan this code and claim your mint Kudos. Um you will be able to claim your Kudos for up to one hour um starting in about um ten seconds. Um and yeah uh claim it and um we will see you here back next week um with another episode. Um and with that just want to wish everybody a very happy Friday and happy weekend and to enjoy uh go outside touch some grass enjoy ETHCC if you're there. And we'll catch you back here next week like I said. All right gang, all the best.

Yeah thank you guys so much for joining and for sharing in ChainBrain with me.