DevNTell - Building Verifiable Onchain Data for Institutions with Edge & Node CEO Rodrigo Coelho

Transcript

Intro

GM GM, welcome to what's going to be another fantastic episode of Däv Talks. So, if you didn't know, Däv Talks is a 30-minute podcast held every week allowing founders, hackers, and anyone in between the opportunity to come on the show and showcase what they built. And today, I'm ecstatic to welcome Rodrigo Cuello, who is the CEO of Edge & Node. So, if you don't know, Edge & Node is the company behind The Graph, and they are building Amp, which is infrastructure that transforms blockchain activity into structured, verifiable datasets queryable by SQL. So, if you stick around for today's episode, you'll get to meet Rodrigo, see a live demo of Amp, and learn how you can get started using it today.

All right, let's get into >> [music] [music] >> GM GM, welcome to the show, Rodrigo. I'm ecstatic to have you on today, man. >> GM GM, happy to be here. >> Excellent, excellent. Uh yeah, I'm really excited for our audience to get a chance to to see you go over kind of the origin story of Edge & Node. Um I know you guys are got really popular for The Graph.

Graph Day Memories

My actual first crypto conference was Graph Day, I think back in 2022. >> No way, you were there. Wow. >> I was there, yeah. >> Oh, wow. >> It was um it was actually it was actually so good that I haven't been to another crypto conference since cuz kind of yeah. >> You set the bar so high. >> Set the bar, yeah. I I didn't want to experience anything else. >> we cannot be topped. >> That's right. >> Oh, that's so cool. >> Yeah, yeah, for sure. >> a small that was a small venue. It was like a theater and >> Um yeah, it was a lot of people, so that's cool you were there. Wow.

>> Yeah, yeah, yeah. I participated >> Was it 18 or 19? >> I think it was um 22, 22 or 23. Um I was there. Um but yeah, I participated in the hackathon uh some prize money there. >> Oh, you went to the one you went to the one at the Palace of Fine Arts, you mean? >> That's the one. Yeah. >> Okay, yeah, there was a prior one that was even smaller than that, like the original one. But yeah, that the Palace of Fine Arts, that was amazing. Yeah, that was like >> a really big venue and yeah, the hackathon.

>> It was a good time. >> Oh, amazing. >> Yeah. >> Cool. >> Um but yeah, that that kind of aside, yeah, it'd be awesome if you just want to give a introduction about yourself before I get into all the all the great stuff you have prepared. >> Yeah. Um Well, a little bit about me, I've been a tech serial entrepreneur for a couple of decades or more and started, you know, early days of the internet, kind of the dot com boom. Founded, you know, a few tech companies and um around 2016 era, started to get interested in crypto and out of kind of right place, right time, I was working at a co-working space in

Origin Story of Edge & Node

San Francisco and happened to become friends with Yaniv Tal, who was like the original founder of The Graph. And we he was actually working on the pitch deck for The Graph at that time and I was kind of fellow entrepreneur, kind of like, "Hey, what are you working on?" He had like an Ethereum sweatshirt on and so we started talking. So, he showed me the pitch deck and I saw it was a really big idea and just wanted to be helpful as a fellow entrepreneur and friend and so made introductions um to investors and um and then yeah, a few months later, I was kind of looking for the next thing to do and the timing was such that they had just gotten their first check in and so he'd

asked me to to come join the team. So, I ended up being the first hire um back in 2018. So, I was there from the very beginning. Um I was there were three co-founders. I was kind of like an adjacent ops person at the beginning. I was like you know, helped out with like community like with Graph DAO plan that I was doing kind of all the things as we scaled and um yeah, so we we built the Graph protocol um and launched it in 2021. So, it took about 3 years of building and um launched a decentralized network that does indexing of blockchain data and allows developers to query it.

Building The Graph

Um I'm I'm not sure how much the audience knows, but um yeah, it's thegraph.com. There was a token uh called GRT. It's still in use today. We've served the infrastructure served trillions of queries over the years and still serving uh billions of queries per month and it acts as a major part of the back-end infrastructure of many um protocols that are out there today like DeFi protocols yeah, was is really a a key piece of infrastructure that is still utilized today. And so, it's been a a really fun journey. I can't believe it's been 8 years, but um

yeah, so Edge & Node is the original team. Uh the company originally started as Graph Protocol Inc and then we rebranded our name to Edge & Node in 21 when we launched the Graph and the network. And then um yeah, so we've been you know, working all these years in improving the protocol and then more recently we've shifted our attention to more of the enterprise use case uh with our product Amp that's um working uh with directly with like financial institutions and like getting blockchain data into their their uh organizations.

>> Amazing. Amazing. And uh yeah, I don't know if a lot of people realize, but um indexing blockchain data at scale is is very difficult. Not every chain behaves the same and you have to make provisions on on how to kind of adjust for that. So, um yeah, kudos to you guys for for taking on that that early early challenge Uh kind of being the first first movers there. I guess from being part of it since like day one, is there any particular like big challenges or or anything like that that you remember kind of like going through and you want to talk about?

Early Challenges Scaling

I'm sure there was there was Yeah. >> I mean, there's too many to name, right? Yeah, I mean, everything from you know, back in that era there was like the specter of like SEC regulation and operating within like these guidelines and frameworks that we weren't even um you know, weren't quite clear. And so the the state of the art back then was to set up like you know, a Cayman Islands foundation and you know, launch the token out of there. And so yeah, there was those kind of challenges. We were we always operated under strict guidelines in terms of like you know, working with

non-US people, for example, and everything was above board. So there there were like um challenges there. And then yeah, like scaling challenges. So you know, it really took off in usage immediately and just like keeping the infrastructure up and running during that time was you know, incredible challenges. Team scaling challenges like you know, it's part of the startup game is it's there seem to be like every day another problem solved. So yeah, we've been through yeah, quite a number of them over the years.

>> Yeah, it it yeah, it's just the name of the game. Just how startups work. And you mentioned it AMP being one of your the latest products from from Edge Node. Kind of looking through it, it feels kind of like a evolution of what the graph was initially. At least from from my lens, um, but I yeah, I'd love to hear more about it, um, just what it is exactly, why it was built. Yeah, so we we originally built, um, subgraphs on the graph network, which was like, you know, a decentralized network of nodes that are constantly monitoring different blockchains, indexing the data in, which is sort of like, uh, filtering and, um, storing the

Why Amp Exists

data in a database format. And so, what we started to see was that financial institutions were coming on to utilize blockchain data, like in, you know, 2021, 22, starting to experiment. Our code was open source, and, um, yeah, we heard through the grapevine that like large financial institutions had downloaded the software and were running it internally. You know, it's like, it was MIT license, free to use. So, they were using it, um, inside their organizations. And so, as we started to do more, uh, R&D, we started to see that what we had built wasn't was kind of purpose-built for some particular like developer use case.

And these larger financial institutions had a different use case, which is like, they have existing data pipelines, they have, uh, data analytics teams, they're they're wanting to bring data into existing systems like Snowflake or BigQuery or like big analytics platforms. And so, uh, yeah, we spent a couple years in R&D kind of building the state of the art in terms of uh, being able to index, extract, um, the blockchain data from different blockchains to keep up with the chain itself and handle all the kind of complexity around dealing with blockchain data, and then making a

simple, um, tool and platform that like data engineers could just use SQL to like filter the data down, create data sets, and then pipe it into uh plug into existing systems. It might be an Oracle database or it might be a Postgres database or it might be a big query or Snowflake. So, you can see it um as a pipeline. So, view it as a as a means to like extract data from the different blockchains and then pipe it into where it needs to go within an organization.

>> Amazing. Uh yeah, I I'd love to love to see it in action. Uh as I understand you got a you got a little demo demo for us. >> Yes. Is it uh shall we switch over to that? >> Yeah, let's switch over to that. Uh and then uh I can ask some uh some of my other questions kind of while you're doing that. >> Okay, great. Let's do it. All right. So, this is actually a preview of um what's forthcoming. So, AMP itself is really you can utilize via like um command line or via a desktop, you know, like a data engineer would would be utilizing it

Amp Cloud Demo Walkthrough

more in like a text- based interface. This is um something we're calling AMP Cloud. So, it's yet to be released. So, I'm giving kind of your audience a preview of what's forthcoming. But basically, um this is a a UI visual interface to view and interact with data. And you don't have to be a data engineer to do it. So, we start with kind of this natural language search for like what data you want to look for. So, here we can say like show me like USDC transfers on Base and Ethereum and the LLM uh determines um what tables it needs to look at and writes the SQL query.

And so, it generates the query. Uh you can copy it, you can edit it, um and then view the data that you want to see. So, it gives you a preview of kind of what data you're looking at. So, here we're looking at, you know, base and Ethereum. And again, we're looking at ERC-20 transfers. So, it's a very simple interface, but it's doing highly complex things behind the scenes because obviously it's keeping up with two different chains. You're now joining that data, uh, using a SQL statement. You're getting it stored into a table. And then, um, from this you can then save it as a data set. And so, you can think of a data set as like a table, like a database table that's constantly, uh, keeping up to date. So, every new

block, uh, a new row is added if the if the data matches the criteria you're looking for. Uh, and so, this gives you a preview of, saving the data set. You can look at the the the SQL, of course, and modify it that way. Um, and so, you can create a data set. And how this is how you'd query it. Um, if you export, it will, um, give you the information as far as like, where it's going to go. You can run it once or keep it continuously running.

So, keeping it syncing block by block. and reorg. So, reorgs are kind of like where a chain has to go backwards and then go to like, uh, an uncle or a fork. And so, sometimes that happens periodically with chains where it needs to go back and then goes down a different path. And so, Amp handles all of that, too. And that's one of, uh, a major complexity in dealing with blockchain data because you might have had exported data, but then if the the chain went through a reorg, it had to go backwards and then it's on a new chain and now you're the data you had before is no longer in sync.

Verification and Reorg Handling

And that creates a lot of problems. So, if you're dealing with, um, for example, with like, financial institutions that are doing like, um, needing accurate data for like, to make trading decisions, You can't have uh data be inaccurate and not match uh what's on chain. Uh so, that's a huge benefit to um what we do with uh AMP. And so, yeah, you can export this. It creates a pipeline. Um and so, like the pipeline, this is a cool feature. It gives you a visual representation. I think the audience can help uh see this here or this helps visualize kind of what AMP is doing, right? Cuz you have a source

which on the left is like it's coming from um two different places. It gets goes through a transformation. And then it gets sent out. Uh and on the right here is the destination is like an uh big query database, like a big data warehouse. And so, you can create as many pipelines as you want. Um and so, like if we look here uh back at the beginning, uh these are like data sets that you can create. And so, derived data sets are ones that I've created myself. And so, if I log in here, I can see um raw data sets is like all the data.

Derived data sets are data sets I created using either SQL or I can use AI and it creates a uh subset of table a table of information that I want to look at that's constantly uh keeping up to date. I'm sorry. I misspoke. Your data sets are ones I've created. Derived data sets are like community based. So, actually like within your organization or you can have contributors that are creating data sets. And so, it may there may be a data set that that was already created by someone. And so, you don't need to necessarily reinvent the wheel and and create your own data set.

And so, um yeah, you would see like uh the data in terms of what's there. Uh you can see pipelines. And you know, see the ingestion rate and see what is going on there. And then uh this is where you can look at like SQL statements themselves and kind of run a query. And uh one cool thing here is this verified Um so this verified stamp. So, uh the verification piece is a big one.

So, this is where um something that AMP offers is a cryptographic proof that the data matches what's on chain and it uses like the hash proofs that uh you know, that it matches like block headers, etc. And the Merkel roots of the the blockchain. So, um so this So, for an audit purpose or a compliance purpose, like um someone uh like in an audit team within a big organization could come here and spot check or pull a report that verifies that, you know, the data that is being looked at

for, you know, or the data sitting in the database or the data warehouse is is fully verified and matches what's on chain. So, that's like a uh the ability for, you know, audit teams or compliance teams to make sure that the data at rest, you know, is it's a sort of garbage in, garbage out thing. And so, it ensures that the data is correct and complete, which is a big problem with blockchain data because nodes can drop data. Um you might be missing like a block, for example, out of say it's every 100,000 or uh but even then, like when you're s- summing a total and you're trying to get a total and if it's off and say there

was like a large transfer in a block that was missing, your number is not going to match. And so, you know, for financial information, it's got to be 100% accurate, complete. yeah, this is um one of the another cool feature about it. So, really this is like a visual representation of what you can do and sort of like a command line for AMP and most data engineers would utilize like, you know, kind of notebooks in Python and and what um that interface. We've added this um that's forthcoming so people can start to play with it, see what they can do as a stepping stone towards moving AMP internally. And one of the main

Enterprise vs Cloud Offering

differences with AMP is that this is this is run as a like an on-premise license software. So, So, AMP Cloud will be something that you could um you know, create a login and like um leverage it like on a monthly basis, kind of like a SaaS, yep, which is yet to be launched. But, AMP itself as the engine is designed to operate as like a license software. Like you'd license Oracle, you know, and you basically run the code within your secure um cloud environment, which is important for these like larger financial institutions that have like highly um secure uh data needs, that have highly secure infrastructure,

and have uh existing data engineering teams. And again, like they're running um data lakes or data warehouses in-house already. So, this is where um it's an enterprise that's gotten to a scale where um data can you know, paying for data on a consumption model is getting very expensive. And it's like where the when they ask the build versus buy question, like um do we want to build what AMP does ourselves or can we just bring this and and use it um as part of our data infrastructure stack, then we'll do this.

Um also though, AMP Cloud what I'm demoing is is meant for like maybe smaller users. And so, um you'd be able to create a login, you'd be able to have access to it at like a on a monthly basis at like a lower utilization point and then you could start to export data into a database or a system that you might be running in the cloud for yourself.

So it's meant to be a sort of like a a stepping stone. So you might not be ready for like running and even large institutions, many of them go into like POCs first or like we'll run it. So and then do a full implementation to bring it in house. So this is also meant to be able to get up and running like time to value is immediate so that you can get access to the data immediately get value and then start utilizing it and then maybe you're preparing for like a larger project where you're going to like install and run it on premise. So Yeah, that's amp amp cloud that's forthcoming and we're

really excited to to share with the world. >> Yeah, and then thank you so much for for sharing that alpha with us in the audience today. We love we love our alpha on this show and yeah, it looks really neat and I don't know if a lot of the audience caught because like a lot of the audience is technical but very suddenly like you you guys basically transformed some you transformed like blockchain data that like blockchain engineers basically mostly understand into sequel that most like most engineers regardless of whether you're web two or web three can understand and query and that that's that's a very powerful construct.

Why SQL and AI Native

Yeah, do you want to just talk talk about the the thinking and decision making that kind of went into that? >> Yeah, so um the graph itself and subgraphs were built on top of GraphQL, which was um a language utilized for more like web application access. You know, Facebook developed it and it was an improvement off of like using REST APIs. And then, yeah, we went with SQL because, yeah, it's it's kind of the most broadly known language in the world, I think.

that most everyone uh knows and understands. And again, it allows you to like write a simple uh simple SQL statement to um form a table what you're already familiar with of like blockchain data and then store it in a table format. And then also, um AIs natively speak SQL. So, that's another benefit. So, when you connect you ask natural language questions like that, um the AI itself can write the SQL statement, understand the tables very easily, and then uh generate uh the code for you that you need to to look at that data and uh do analysis on it.

>> Absolutely. Yeah, and um yeah, I I know there there's a lot of folks who um know know SQL at kind of like a um like a beginner intermediate level, but then when you want to really like get into the DBA kind of like joins and and and whatnot, like it gets it can get pretty pretty nasty when you're trying to join multiple tables. So, um >> That's the cool thing about AI now is you can just ask it, right? And it will write the join statements for you. So, you can say, "Hey, and then you get the table preview and you're like, "Oh, I didn't want that. Could you show Could you add a column that, you know, grabs this data from this table?" And then it'll write the join for you. So, that's the great thing about AI now is you you can uh know enough to be dangerous,

but then, you know, the AI can like enhance your your SQL writing capabilities way beyond. >> Absolutely. Absolutely. Uh and um what you demoed um at the this um the utilization of AI and amp data together. Is that also available programmatically or are you guys just planning to have Yeah. >> Yeah, it will be Yeah, I think there's like MCP server access and um So yeah, you can point LLMs at it and yeah, do analysis on uh the data itself.

Um so yeah, it's really built to be AI native and compatible. >> Excellent. um yeah, as we know like AI is AI is on everyone's timeline like regardless of what industry you're working in. Um but agents themselves might want to query this data as well. Um are there any plans to say make it available for paper requests via like X 402 or some other payment protocol?

Agents Payments

>> Yeah, potentially um once we launch amp cloud, we're we are looking at all kind for models there. So that already I think that already exists on the graph itself. Like they created a graph MCP. And so yeah, we'd likely follow suit with um with amp cloud. Uh yeah, we were very early on the X 402 path um we we actually started researching micro payments back in 2021 for the graph itself because as you imagine you're doing billions of queries and you're paying sub cent amounts for each query. And so we we saw okay, paying the

gas fees for each query wasn't going to make sense cuz it was like the gas itself was more than the query itself. So we looked into micro payments and developed like a a batching system with that's utilized within the graph. And so yeah, we were researching agents in late 2024 and then um and how agents might pay. And then X 402 came out. And so we we were early adopters of X42 and contributed uh with the Coinbase team and met with the Google team and with their work with A2A.

Um yeah, and we developed a project called Ampersand that was like uh that utilizes amp behind the scenes actually. And so it it's uh a platform for observability and control over agentic uh commerce and payments with like the ability to have like uh a rules and policy engine and all of that. So, yeah, AI we're very AI forward. and yeah, we've obviously leveraged AI in the development.

And um I did a talk recently at like a data data management summit here in New York and was specifically about how um AI is creating like a massive uh amount new data. And so like the amount of data is increasing. So like storage costs go up, but at the same time the ability to uh create kind of bespoke or um purpose-built data infrastructure uh that can cut your cost dramatically offsets that. And so we did that ourselves at EdgeNode. We cut like our cloud bills by like 84% over time. And amp is kind of the first

example of like um a purpose-built uh piece of data infrastructure. We believe that's that's going to be a mega trend going forward is that now AI uh lowers the cost of developing um kind of custom infrastructure. And so you can have the AI you work with the AI and it will give you feedback on what tradeoffs you can make. So, you know, I the AI might say you can cut costs by like 80% but your latency is going to go up by 10% and you decide, "Okay, I'm willing to make that trade-off if we make this change here and you kind of go through that iterative process.

And so if you're looking at like a large enterprise and they have like an increasing cost of data because of what AI's are producing like more logs, more transactions, and more storage costs but then the data infrastructure costs fall what that means is on the same budget like say the annual IT budget or AI and data and infrastructure budgets like say 10 million. Um say they've cut their costs by like 75% on infrastructure but the cost of increase here what that leaves is more there's on the same budget there's more ability to spend on AI and tokens and

everything else. So by leveraging and pulling levers you can sort of find more budget for like additional AI and you know lower costs on what you're spending on like cloud bills or yeah like costs that really could be cut. >> Yeah absolutely I I'm I drink the same Kool-Aid and yeah absolutely it's it's it's some wild times we're living in but yeah I'm I'm here for it. >> Just before we let you go as we're approaching time just wanted to give you an opportunity to to share how people can get started with Amp if they're interested in

How to Get Started

playing around with it perhaps if there's like a beta program they can sign up for to to play around with what you demoed. >> Yeah the best way is go to edgenode.com um it'll probably be linked with the show but yeah if you go there there's just a talk to us button yeah we can set up a demo um and then once yeah Amp Cloud is out there'll be like kind of a self-serve option where you can like register and have like a free tier and play with it. Uh but for the time being, for those that are interested, best way is um reach out to us via the website. And yeah, we'll we'll set up a call and walk through everything and show you how it all works.

>> Amazing. And yes, indeed, all those resources will be in the description of the uh podcast below. Um so, definitely check that out if that is something that interests you. Um and with that, uh Rodrigo, I just want to thank you so much again for taking the time out of your busy day to come chat with us. Um yeah, I I I think what you guys are building is really cool and yeah, wish you guys the best of luck as you as you continue to to build it out and release it to the wild. >> Fantastic. Appreciate the time. Yeah, my pleasure being here and um happy to have this chat with you. >> Amazing. Amazing. Likewise. And uh with that, I just want to wish everybody a very happy Thursday uh or or uh Friday, wherever you may be, and that we'll catch you back here tomorrow, actually,

Outro

for another great episode of Dev and Tell. Till then, have a good one, folks. Cheers.