Founder + VC + AI = Decacorn

[Music] being with us today Elia thank you for joining us I'm excited about this conversation because Elia has many insights to share being a VC fund manager who are actually using AI in their investment thesis and in their investment process they're also investing in some of the most disruptive AI companies around the world and as we're going to hear from Elia today they have compiled one of the larg EST data sets on Startup and VC activity um that can give us some insights into the latest trends in investing and investing in AI so Elia maybe a quick introduction for the audience about Quantum Light and what you guys have built sure so uh we are a quantitative Aid driven venture capital and growth Equity Firm um we are uh we were founded by a guy called Nick Stony who is the founder of revolute and my business partner in this venture um we are a team of 25 Engineers data scientists Quant Traders um investing our fund uh and uh you know what this means is that we use uh models AI um rather than kind of traditional VC methods to find and invest in our in the companies for our portfolio what what does it exactly mean you use AI can you be more specific and maybe for the non-technical you know uh members of the audience explain it a little bit more sure so uh we collect a a vast um amount of data so we we currently have I think just 700 million LinkedIn profiles each one has at least six positions um we partner with scores of data vendors uh then we aggregate this data so we build pipelines to them and then we engage in what's called data curation which is a very common thing to do in quantitative investing uh and so we build a kind of proprietary data set um and then we start to analyze to see what kind of factors could have predicted performance of companies over time uh be that um you know biographical features of the founders uh who the other investors are what kind of category the company is operating in etc etc I think we analyzed over 300 different features to find about 40 which um work now when it comes to AI how how we're using that so of the text there there's a vast amount of text information about startups and Venture back companies and so uh we we process that to uh find signals as well I'm trying to imagine this data set you're talking about so LinkedIn makes sense to me it's easy to imagine that you've bought this data and then what other inputs have you put into that data set yeah the so the vendors we partner with they range from kind of the more traditional ones like kind of pitchbooks crunch bases CBA insights of this world uh through to scrape ERS like uh which can generate LinkedIn reviews uh web traffic and so on uh through to manual labeling so there's a huge industry of uh manual labelers in the in the AI space where we can construct our own uh features curious how big is it um I mean it's big data so uh yeah billions and billions of of data points uh so um you know a lot of parallel kind of processing to evaluate it from my own experience you know whatever I looked a company I'm personally either invested in or familiar with on these large data sets I find a big degree of discrepancy with what the valuation is online and what the real valuation was how do you deal with this type of discrepancy yeah data curation you need to understand exactly how the data is generated uh what to trust what not to trust you have to fill in blanks with models uh where you can so um yeah it is it is a a tougher challenge than in public markets but uh we certainly believe it's feasible we have 20 years of LinkedIn data for example there is a huge industry of manual labeling that you can uh quickly create a data set so uh we certainly think that in the future most Venture firms will be doing what we're doing so now you've built this data set what's next what's the next step well uh so it it basically um you know we use a model uh like an AI Tool uh that tells us whether to invest or not in the company which is where the unique kind of angle uh happens uh I mean Venture capitalists say that it's all about pattern recognition and a uh systematic you know model has Perfect Memory it's not swayed by emotions uh it you know it it knows everything uh so it's certainly going to perform much better than human um you know certainly you know our models suggest that uh we should be aiming for returns of well greater than 2x top uh top desile or top 5% we see returns so over 10 X on the whole fund so I'm a let's say I'm a Founder I want to get money from your VC fund do I apply or do you guys reach out how does it work exactly yeah so we reach out uh yeah so we're purely outbound business because we we have a list of companies that our system thinks are the best uh Highest Potential in the world uh and so yeah we we we come to you and uh then you share some data with us we update the model and if it says yes then we proceed to invest so the process is super fast and very so it's more of a radar to pick up the most promising startups around the world then you reach out to those Founders um What stages and how does access work because we all know like VC is a game of access as well yeah of course so um yeah we we reach out to Founders it's typically at series B stage because um you know if we're using a data driven approach uh there has to be good data about the company there has to be a few turns of the head count for example they need to be other investors in the in the company to understand uh figure out the signal um I mean on the on the side of uh access uh I mean my co-founder um nick stonky uh is one of the best entrepreneurs uh in the world you know I have an entrepreneurial background um and we basically share the Playbook of how to build the next revolute you know how to handle hiring how to handle uh International expansion uh how to do Performance Management so you can just copypaste this and radically accelerate how grow your company um and that's what gets us access in the deals I've heard from a lot of VC managers that say they're using AI um I wonder a how big is your fund are you just investing investor money or are you actually betting your own it's it's it's not a question usually asked on stage about GP contributions but I think you know with such an Innovative investment strategy one should ask if it's all P money or is it is it your own money that you're putting at risk here yeah uh so we're are 200 million fund and of that 60 million is our own money so we have like 30% GP contribution which is probably one the highest in the world uh for for this type of uh Venture so you're betting $60 million of your own money on a Model you have built using a 24 person engineering team um and as I understood from you before the session some of these Engineers or most of them maybe are in the region here some of them are yeah uh I mean we we have a very high standard for hiring Engineers uh like we hire people who like want International physics maths olympiads uh data science competitions uh we also have a very strong Quant Trader team that evaluate alphas and Elia you feel this will outperform the the VC bench marks yeah absolutely so I think if you think about the VC um how how VCS evaluate uh what to invest in uh they usually weigh up a large a largish number of factors so like who are the other investors in the business what their track record has been uh what's the category and what what sort of like amount of investor dollars it's attracting what is the team like uh what skills are on the team um sort of what's the um valuation increase over time and they're trying to pattern match that to um you know historical successes or failures and trying to find uh sort of situations that are similar and and I think uh you know there's no good reason why a machine shouldn't do this better and faster and also protect against these cases where something's really overhyped um you know one thing we learned actually is that the degree to which a deal is over subscribed has nothing to do with the onward prospects of the company so it's actually like hot deals or not they have about equal chance of success which is a surprising result right so uh and generally speaking in VC there's never a case that uh you know oh it's a great company um you know strong founder team huge growth amazing category exceptional investors and they're willing to give it away for free there's always some risk in in in this in any investment and and um you know having a system really helps you really weigh up those those factors in a way more precise uh way than than a human possibly could so um wondering for example a company like FTX you know amazing investors seya one of the best names in the world writing a $375 million check for me this is like a great signal um did you test it through your and what would what would your model say um yeah so what knocks out FTX why we would not invest in it is it's actually the the increase in valuation between series a and series B rounds which is an important feature so a kind of sustainable pattern of increase in valuation is let's say three to five times times every sort of three and a half to four years uh now I think with some of those stories uh you know FTX were I mean there were other I mean FTX turned out to be a fraud but you know like Clubhouse and and I mean I I can find more uh but uh really there was a huge gap up in in price um so so our system would down downgrade that what have been some counterintuitive insights you uh sort of learned from the system so far um count intuitive uh I mean yeah there there is there is a bunch of in I would say uh really the the interesting thing is how um it's really a combination of factors that can be really different uh in terms of you know finding the best uh companies um I mean we also occasionally find like Curious factors so for example uh founder features we tested whether you know stem education or not expat or not age um so um we tested whether even Dropout from college predicts um performance and the answer to that is that it only works in combination with top university so if you're going to drop out you better drop out of Harvard um so yeah if you're in a kind of mediocre University by God you should finish the degree you you have have uh maybe maybe you can tell us uh also a little bit about where you are in this journey are you did you start deploying how did the first you know inv M SCH if you did yeah absolutely so uh we we started deploying in about June this year so uh we we already are like about 20% deployed so our thesis on access is clearly working like we're able to get in virtually every deal that we look at um we uh we invest a lot in AI as well we have we have a unique understanding of AI where you know our thesis is to find companies with application space uh with unique data sets very strong teams that uh you know are about to inflect uh and we were able to find those in the medical space for example so we're investors in a company called radii and deep scribe which are uh sort of emerging as leaders in uh all kind of medical note transcription um in in the US um we we invest in legal uh Tech uh AI companies and that's just to name a few so my final question for you for the investors and the founders in the room you seem to be purely outbound your model is identifying these winners and then you're reaching out to them and you're getting that access do you feel that within five or 10 years this is how VC is going to operate and for the founders they should maybe prepare for a world where don't call us we will call you yeah I think it's a big market so you can't say that everyone will operate this way I think where we are right now is like hedge funds were in the 70s if you remember in the 70s the hedge funds they were driven by research so uh you know people were reading annual reports and creating a model and you know analyzing fundamental data then uh the Advent of Quant uh investing really be came and some of the Quant hedge funds have dramatically outperformed these kind of fundamental strategies so some produced the best returns in the history of returns in VC we're a bit still like hedge funds in the 70s and we're still using kind of uh partial information and and and and uh you know uh research type reason to find to find companies for sure quants will be a large share of the market in the next 10 years um no doubt about that in my view Elia thank you so much for sharing these insights with us and of course best of luck with the with the fund thank you thank [Applause] you