Efforts to create transparency in DeFi through statistics and ML - Marijo Radman | Solity Network

does it work let's see yes perfect okay ladies and gentlemen I believe uh most of you here in the room can agree with me that D5 is still a very dangerous place to be in even if we just look at the recent events from the last summer starting with the Terra Luna crash followed by the FTX crash happened last November the usdc Deepak in connection to Silicon Valley Bank and even recently with the Euler hack we should ask ourselves why is this happening um I mean there are these companies in the space who are dealing with the security on the protocol level like auditing companies there are companies who are dealing with economic incentive designs on the protocol level like call slabs and Gauntlet and yet we have more or less a systematic repeat of these events like even we cannot talk really about black swans because it's not an unexpected event if every few months something like that happens and the question is actually okay what is happening there I mean we have these companies they're analyzing that yet these protocols here they are just old few years like two to five years and yet they're managing to attract billions in Assets in capital and if D5 wants to go mainstream D5 has also taught Tech that side of the uh of the plate uh in the end also liquidity provision that means like this one who provide the capital have to know what kind of risk they are dealing with and I mean just to Summertime when the device movement started we promised ourselves okay we are going to build a new Financial system which is going to be open transparent more secure without a centralized entity in the end what did we achieve we have now uh Financial system which is actually worse than a traditional one because those kind of events are not not happening every few months there um just figured what's about myself I'm the CTO and co-founder Authority Network spent most of my career working for major German corporations like Audi aeon said f um I'm in crypto since 2016 mostly working on the D5 side and salty network was co-founded beginning of the last year we closed our first financing round in September of last year uh since then we managed to build a team of 18 people out of which 14 are working on the development side of solity and besides that that we are headquartered in Munich we have also offices in London and Zagreb to actually explain uh the main problem of D5 the moment let's just take a simple case here all of you probably know this uh popular strategies that you have your stable coin and you put it in some of the kind of the projects and then you're yielding additional interest on top of that and uh basically increasing your Capital here a simple case we have oven compound what is the way basically to decide where to invest your asset at the moment currently what everyone is more or less doing at the moment is they are checking the yields on one side of the liquidity pools they are checking the yields on the other side and just moving the Capital there I guess most of you can agree with me this is not really the smartest way because you're not taking the risk in account a better way would be actually to go in a direction of risk adjusted returns so basically assessing the risks looking at the opportunity and then moving your capital the problem is like Ave is basically built on top of a chain like ethereum between ethereum and Avatar is usually also L2 like arbitrum if there is a other ecosystem for example Avalanche then you have subnets and actually now you'll see that D5 is nothing else in a composite of various Primitives and the thing is like all these getting even more complicated we have now over 100 protocols we have Bridges we have these yield optimizers yield aggregators and for a proper working or for example lending protocol you need also oracles and all of them are in the interconnected that means always connected through liquidity to compound if something happens on the Euler hack you will see it also on the other side and that's basically the main problem how to assess the risks here all these Primitives there when you are trying to deploy your Capital you have to deal with all this stuff here it starts with the chain you have the consensus uh you have the validators you have their governance structures then you have the l2s with their governance structure the sequencer comes into the play you have the D5 protocols it just starts with the security side of smart contracts but also the liquidity pools with all these liquidation events that are likely Duty pools where you get inside you are not sure you are able to get outside on top of that there are these yield Optimizer seal allegators they are bringing additional Securities issues additional centralization and then it's getting even more complicated when you have cross chain strategies then you have to deal also with Bridges and then you have the finality of one chain the finality of the second chain and you have the bridge which makes actually everything worse the recent Innovation discussions we have on the consensus level which shares security risk taking is going to magnify the whole topic even more I mean I'm not talking here about just oracles and chain link where the data consistency can break and then you have an attack on a D5 protocol once we're going to share this security the definition of systemic risk is going to even be more relevant and as you see here now if RV breaks you will have consequences so compound if something happens on arbit room you will feel it on Ave if something happens on more for for example you will see it on spool and so on and now the question is like okay how to deal with that um this is actually the main problem like the when we look at D5 a new risk category was introduced is the fundamental side of the risk it's not known in traditional Finance space and the current approach is how to deal with that for liquidity provision are not suitable for that they are at the moment subjectives that means like you have a team who built some kind of model how to assess the risk according to their domain expertise they are simple so basically it's a very simplified model so basically what I saw for example for smart contract security they just check if there is an audit uh how long is the Smart contract deployed how much TV do the SMART monitor have so basically with more time it's more secure this is the assessment of a smart contract and then lastly they are static like all these Primitives I showed you before even the smart contracts are usually not changeable all the other elements are Dynamic and they are constantly changing like and these scores are not taking this in account and that's basically where we at solid networks came in we are deploying at Big Data approach so basically by indexing and monitoring the whole default ecosystem starting with the L1 L2 chain like for example ethereum or arbitrum D5 protocols stable coins Bridges oracles including the liquidity pools focuses on the fundamental side of the risk so we are getting data on security liquidity flows decentralization governance socials and development activity and then enabling our customers to build risk related products which can be customizable risk costs which updating in real time it could be like more advanced risk screeners for insurance protocols but also alerting and monitoring system and we are also aware of that if we want to enable the next generation of D5 which is going to be like smart we have to think about the centralization the centralizing the whole text stack from Storage from Computing up to delivery to the protocol uh when we look at the space at the moment what's available there so basically that's our definition we have the protocol design level and we have the liquidity provision on the protocol design level there are many companies out there like from the auditing companies like certic up to the economic incentive design companies like our slabs dial routing systems like fota yet on the liquidity provision side not really um the econometrical side is done by a few companies usually by teams who are coming from traditional Finance space and they are approach uh applying econometrical models to deal with the risk on that side yet on the fundamental side of the risk so a broader approach to the risk the companies there are what we're in that file like most like exponential and device safety and that's basically where we had solid position ourselves the thing about the solution what we aim to do it's very how to say it's like complex like you need a lot of domain expertise and we are aware of that to deal with all these facets there are many companies and that's the reason why we said okay let's build the risk infrastructure let's enable uh the next generation of D5 by building a D5 validator layer so basically we are at the moment running our own notes to get the on-chain data and we build the whole data pipeline to get the off-chain data we are also building in the moment this infrastructure routing layer what we call it like this and the idea is like to onboard various middleware companies in the space from the centralized storage up to decentralized computing um with various versions why we learned in discussion with many customers that they are all in a different stage of decentralization they understand the centralization or very very differently and even getting more complicated when you have to explain them that in the end you have to pay for the centralization and that's the reason why we said okay let's onboard various Solutions there let's create a routing various routing options and then incentivize the routing options which is going to be mainstream so basically that companies who are fine with centralized immutable databases a normal centralized Computing and Oracle Networks there are also startups who wants to have everything decentralized and they are willing to pay for that and we are going to enable that and then on top of that basically enabling the processing running various algorithms from statistical analysis network analysis machine learning deep learning and so on and the idea is basically to build a customizable risk products it could be like a risk cost for wall strategy rebalancing up to portfolio rebasement it could be more advanced risk famous for insurance protocols liquid staking or just alerting system for for example a trading infrastructure if there is in place or for tradeify or a new protocol which is deployed the delivery of course apis for the companies who are comfortable with that or through oracles if we are talking more about D5 protocols the customer groups we are targeting the moment are really the liquidity provision oriented companies from yield aggregators yield optimizers liquids liquid staking Insurance protocols but also Landing protocols I mean this is the current stage what we are currently building but the idea is actually something else opening up the platform enabling third parties provide us to provide data which is not currently available there enabling third parties to provide modeling power modeling Solutions which we are not currently giving here for example a good case would be like agent-based simulation a modeling which we are not capable to do because we don't have the domain expertise on that side but there are companies who are willing to integrate with us to enable that and also data providers who have for example data on security sites data on liquidity provision to build more advanced risk Frameworks and in the end those risk Frameworks which are going to be used by the protocols in the end they are going to be like paid through that solution and basically they are going to get their shares through that and the idea is like once all these third-part party providers come inside there is enough data and enough variability to build more advanced models which can address the liquidity provision side of that um we learn from experience we saw from other projects when they started with the idea of decentralization and opening up that this was not really a good idea in the beginning to do that because they didn't manage to get immediate value to the End customer and then in the end they managed to get the network effect that's the reason why we said okay let's go with the centralized case let's build the first part by ourselves and then opening up the platform from the modeling site up to the data side and I'm happy to announce over the last eight months we managed to launch our beta solution and here is just a short overview how it's working we made it very simple very intuitive that everyone can create their own risk framework with very little knowledge about risk but if they have domain expertise they can use that for their use case in this case like blocks Bluetooth framework it's from the conference from the last week as you see basically they just select the ecosystem they are moving around out of these various Primitives and they can click save and automatic a default risk model is deployed and are we how we assess the risk is basically by building sub-risk models on various topics from security up to liquidity because we learned you cannot build a big model who analysis everything you have to break it down and then enabling the customer that he can do optimization if he's comfortable with simple cases like changing the waiting factors of all these various models to build his own adjusted solution in the end but also what we want to enable domain expertise people who know who understands defy to go on a parameter level no input level and changing the weighting factors of every single parameters even the modeling side to build something specific for their use case which can be used in their X Epsilon protocol we don't even know at the moment um we started on ethereum site we integrated most of the blue chips and from here on we are now going in the direction of ltos integrating arbitrom and other L2 Solutions and from there on um to other l1's where the liquidities for example Cosmos chain and binance chain uh with the idea to attract as much as possible customer and I'm also happy to announce that we are going from now on in the opening in the centralization of the whole solution that means by the end of the year we are going to have a first running solution which is going to open where third-party Solutions can come provide additional data and they can build additional models which we are not currently addressing the customer or the target customer what we are currently aiming for is yield aggregators we integrated with the first paying customer which is spool.finance it's a integrator on ethereum and also traditional Finance piece mostly hedge fund managers and currently with the beta solution is tested with roughly 30 design partners and lastly just to conclude I believe the quote from Dan Market is still on time um D5 in the current form is going to change and the next generation of D5 is going to have a strong risk management and we believe that solid network is going to enable that so if you have any questions feel free to ask or join just our telegram group and we can continue yes hey so you're collecting data sources from the blockchain right so how can someone be sure that you're not biased you didn't change the data and you're not completely biased on it yeah this is the topic actually the question is like at which stage you are addressing that now we started with a centralized case it was the trade-off we made so basically you have to trust us but in the end you get immediate value you can build risky models once we open up platform there are going to be proofs they are going to be like third-party providers who provides the centralized storage immutable centralized databases with proofs and then you can have the trust as a D5 protocol to to integrate that we are aware of that it's a portrait of at the moment to get as much as possible traction um why is not the graph Network working out for for you guys we started actually working with graph and then we learned that a lot there are a lot of sub graphs who are not standardized there are subgraphs doing the same thing with the different data cases it just starts for example with yields and you will learn like you have three subgraphs given to different yield and how are you able to trust that and if we as a company have to provide centralized Keys providing value then we have to do it right and that's the idea we are just that I would call it First Data provider focused on this scenario case what we have and then later we are going to open up the idea is like to onboard as much as possible security related companies who have amazing data sources from D5 up to go plus that are also on the economic side companies who are doing amazing stuff even block analytica here and the hope is basically to onboard them to help us to bring their domain expertise to make this model even smarter than they are currently any other questions well I just wanted to say uh once that does open up it would be really interesting to use uma's Oracle to be able to pull in the data if it's verifiable and I'm also wondering if the model outputs would be in some way verifiable that there might be some cases that you can like run the model again and check that it ran in like some specified way yes I mean okay there is like you have the Computing side there are like providers we are working with who provide let's call it the centralized Computing networks where you basically have a consensus on on the modeling side you cannot get it constantly right because it's ml side but you have some kind of thresholds and you decide on that side and bring it up we are also very open about Oracle that's basically also the decision from our side we are never going to go in this direction to build by ourselves Oracle so we are going to enable that someone can use chain link Redstone Seda whatever Oracle is out there not only that we are actually going to onboard this Oracle here because of this idea of composite of the risk like of The Primitives that someone has to see the difference between chain link between Seda between Redstone what kind of risk he's taking for example on a more popular data from The Real World for example price and all of them not all of them are same and basically then in the end the user has to decide what he wants to use okay any other questions okay great I would say just last load as you saw our beta version we have 30 design Partners if you want to test our solution in the current form uh like just ping me I'm here at the conference or join our telegram group and we will provide your test accounts we are really looking for feedback to improve the user experience to look at the data sources we have at the moment what you believe should be like also there that once we go in the centralization and open up the platform at least some base threshold of video is there that people are looking at that as useful okay thank you