Coffee chat with Pietro Mondini, Head of New Verticals & APAC Regional Leader at Kellify
Here it is, the second episode of the #FintechInterview, the new format where Bocconi Students FinTech Society will be interviewing leading fintech managers & entrepreneurs.
Our guest today is Pietro Mondini, Head of New Verticals & APAC Regional Leader at Kellify.
Hi Pietro, thanks for being here today with us. Can you please introduce Kellify to our readers?
Thank you for having me here today. In a nutshell, Kellify is a science-centric company born in 2017 to support humans’ emotions with science to boost dating up matches, finding out art & collectibles trends, or getting the best recommendations on Video-on-Demand, Food Delivery or E-commerce platforms in real-time. It is made of a team of 25+ data scientists and product architects, spread across Italy and South Korea, that extract images’ value to make them actionable in several markets through AI.
Could you please explain in what fields Kellify operates?
Sure. Kellify operates in the fintech sector by considering everything that crosses our eyes, even numbers and transactions, as images. We started training our AI upon tons of artworks to assess their worth; from there, we spread our energy to the digital world, where we detect pics to understand those that can better influence our clients’ targets and drive engagement in different segments and markets: e-commerce, dating, video on demand — to maximize their performances and driving sales, or helping brands’ creatives in their decision making.
Having moved to this vast scenario paved the way for a new reading of the diversity issue: our algorithm analyzes if our client is compliant with antibias standards, assessing their online and offline channels, tone and image through detailed reports. At the same time, we tackle insurance and investment sectors, providing wealth managers and insurance companies with meaningful insights that streamline their processes and let decisions happen in a data-driven way.
In terms of competitivity could you please tell us what is your positioning in the sector now? And what is your competitive advantage?
Firstly, I would like to point out that AI is a very complex and young macro area and that there are a lot of players fine-tuning its technologies. In terms of competitiveness, we need to look at the single products we offer, such as image detection. Our technology which evaluates the aesthetic of a photo is unique in the market. As far as I know, some players analyze the most suitable platform to post certain content, the best performing influencer, or the right journal where to post ads, but what we do is special.
Regarding our competitive advantage, we have developed a technology which is very ductile and applicable to many goods and markets. This is strongly linked with our company history: we have explored a huge quantity of assets in a relatively short period, shifting from the mere art market to that of images in its broadest sense — considering a selfie as much as a fine wine or a limited-edition sneaker. At the end of the day, they are images made out of shapes, lines, and shadows. Our AI knows that and can capture the chemistry that drives human preferences.
This is what really sets us apart: the ability to break down emotions and translating it into science to drive human endeavors — from creative impulses to actions, words, or thoughts — perfectly matching the actual image-driven context, hauled by digital environments where pictures weigh more than text. This makes Kellify the game-changer of a new grammar, that of blending science and emotion into one single energy that can interact with different players — be it a video on demand platform, a dating app, a wealth manager, or an insurance firm.
Thank you. Let’s focus on your wealth and insurance segments. Could you please make an example of a product of one of them?
Of course. Considering our insurance app, the product we offer is an AI software, through which the insurer obtains the fair value of an asset, in real-time, only by uploading a picture. This allows us to make a policy in a very short time without the need of an expert.
Regarding art pricing, could you please tell us more? Works of art are usually evaluated at their historical cost, so what information can your algorithm additionally provide?
I would say that we do add value both to the insurance agent and the insured individual. On the one hand, our software allows us to give assets valuations without involving the client. These real-time insights make the insurance company aware of the actual value of the good and eventually to change the policy for the client. On the other hand, the client benefits from this, because if the value of its asset increases, then its refund, in case of damage, grows. We basically act as a substitute of art advisors, simplifying the whole relationship between the client and the insurance company.
So, how could your algorithm be used to invest in illiquid markets?
One of the objectives of our technology is to support people who manage capitals in illiquid markets, like paintings or classic cars. We can give a current valuation, a forecast of the good, and its future liquidity level. The goal is to find the most liquid goods in the pool of possible assets, that could give a return of 10%/20% annually. For instance, selecting them from a range of items that will be on sale in an upcoming auction. As I said before, this product is very useful to wealth managers and private bankers for investing and art lending purposes.
May you add something regarding art lending?
Yes, sure. Art lending means receiving a loan providing pieces of art as collaterals. The biggest market is the US: on average, banks’ valuations are conservative, and they give you 40/50 % of that amount. Banks sometimes find themselves in the situation of having a highly expensive painting and not being able to sell it and this is where we act. We help make a more accurate valuation, allowing to lend clients even 60% of that amount for more liquid goods, eventually increasing their number of transactions and commissions.
Technically speaking, how does Kellify’s algorithm deal with speculative bubbles or unpredictable events? How does it manage the uncertainty of the market?
It is a very complex question. Nowadays there is the belief that AI is a crystal ball made to forecast the future, but it is not like that. AI is something strong and powerful, but it works like the human brain, only handling thousands of processes simultaneously. It is a support of the human intellect, not a replacement. Our sector is complicated: what we do is finding out prices and trajectories of artworks and comparing them with other artworks that have been successful in the past.
About bubbles, we need to take one step back. If the concept is the identification of an artwork value, where our investor wants to spend his money, in the short-medium term our goal is to estimate how much time will be needed to sell it and in which market.
What I am trying to say is that you can have the best tech in the world, but the market remains unpredictable. Our technology is powerful but it must be used by someone who does this job daily to properly function. Sometimes, people ask us: “Can you forecast unpredictable events?”. The answer is no; a broker can have an intuition, but forecasting the future with certainty is impossible.
A few months ago, you were working on trading algorithms, evidence of the fact that your technology can easily shift to new industries. It would be interesting if you could update us on this topic.
Improved deep knowledge for selected neural networks would be one way of making our AI ready to perform under different conditions. In order to understand how different points of view could affect trajectories and evolutions of converging targets, typically situations in which several evaluators express their opinion on a single item: many users and a photo, many auction houses and a piece of art.
An impressive school for this approach is the price target of a stock, representing negotiators’ price targets and thus constitute the best settlement a buyer or a seller can expect to achieve: neural networks are also able measure non-linear, non-compensatory decisions that are involved in price negotiations. But above all this environment shows how to fit a regression model using neural networks to understand a multi-evaluators system.
What we do in the stock market is finding a target price as clean as possible. We perform our analysis on historical data and analysts’ forecasts extrapolating patterns and bias from the market, and using them to formulate the most accurate valuation possible. In simple steps, we select a security, we observe what price forecast is calculated by a given group of investment banks, we analyze investor reactions to this, and we provide our estimate thanks to our algorithms.
We have been doing a lot of tests with exceptional results: 90% of the times, we performed a better valuation of the title than other financial players interested in the security done. In other words, we came closer to the real value of the title comparing to the 10 best performers on that title. This product is completed and has been a great success given our illiquid market background.
Let us focus on the current market conditions: how did you deal with Covid-19?
As a company, we have been able to work very well remotely, and being a tech company full of young talents really helped in this way. Even though the office is still important for the efficient functioning of our systems, we managed to work productively on other imminent projects. In terms of demand, we faced a growth thanks to the need of our clients to evaluate goods remotely and the fact that we operate in a fast-growing environment, especially that of digital.
To conclude, can you tell us how your future will look like?
Our significant technological advancements led us to fresh products for diverse industries thanks to AI, a single treasure that adapts to different angles and that sneaks into the consumer’s mind with the goal of embracing everyone’s deepest feelings and unleashing the aesthetic pleasure hidden behind every image. To reach its aim, Kellify is currently focused on onboarding the best team to fuel science and head towards the evolution of our recommendations. After the Series A round, what lies ahead is the ability to train our algorithm based on targeting each consumer behavior path, beyond delivering objective performances. Our goal is to become the largest brain in tune with the next generation of transversal and pervasive competencies, exactly as electricity or water do: while the first one ignites what surrounds us, the second can reach the thinnest and deepest rifts.
Thank you, Pietro, for having shared this precious information with us. It has been extremely interesting. We will catch up soon for new updates on Kellify.
Bocconi Students Fintech Society
Authors:
