INTERVIEW

Arta Finance, an AI-driven wealth management startup founded by former Google employees, aims to democratize financial strategies once reserved for the ultra-wealthy.

With over $90 million raised from top VCs and 140+ angel investors, including former Google CEO Eric Schmidt, Arta has gained traction among tech professionals from Apple to Stripe.

This month, former UBS CEO Ralph Hamers joined as advisor, and the company launched its patent-pending AI Copilot for enhanced portfolio management outside of the US.

Arta manages hundreds of millions in assets since launching in the U.S. in October 2023 and earlier this month, opened internationally in Singapore.

AI Street spoke with co-founder Chirag Yagnik, whose expertise spans mathematics, quantitative finance, and technology. Our discussion covers:

• Chirag's journey from DRW's autonomous trading to Google Research

• Arta's mission to serve those with $500K-$25 million in assets

• Innovations in robo-advisors, direct indexing, and quant strategies

• LLMs' potential impact on finance over the next decade

• The rapid pace of technological advancement in finance

This interview has been edited for clarity and length.

Tell me about yourself

My background has always been at the intersection of mathematics and finance, which serves me well in my current role. I have a graduate degree in mathematics and studied quantitative finance. My career started in systematic trading at DRW, where I focused on fully autonomous strategies. In my first few years there, we used reinforcement learning to quote markets on futures and equities, and I eventually became a portfolio manager building out latency-sensitive systematic trading efforts. Over time, we also expanded to non-latency-sensitive trading strategies.

During 2014-2015, we developed a system that was fundamentally model/ML driven where 90% of the components were shared between research and production environments. This allowed us to scale rapidly, becoming one of the largest traders in both European and US markets. We were trading between 3% to 30% of the overall average daily volume on major ETFs like SPY, QQQ, VXX, and DIA. This, along with other projects I had the opportunity to work on at DRW, helped me gain a deeper appreciation for what it takes for machine learning models in finance to operate at scale.

Over the years my interest in tech was increasing and that led to me moving to tech as the next part of my professional journey. Specifically, I joined Google Research to work on the area of adaptive learning. This is where I started working closely with Caesar [Sengupta] and Felix [Lin], my co-founders at Arta.

We founded Arta because we felt strongly that you could do pretty much everything on your phone except manage your finances holistically. We wanted to serve users who have between $500K to $25 million in assets or make more than $200K per year - people who aren't served by ultra-high net worth private banks but are beyond the very beginning of their financial journey. Arta was born out of a desire to serve that user, offering services typically only available to ultra-high net worth individuals.

Where did the idea for Arta come from?

The idea for Arta really came from our personal experiences and observations. At Google, we were focused on building products at scale that were either more accessible through technology and machine learning, or entirely new products that couldn't exist without these cutting-edge technologies. 

Waymo is a perfect example - you simply can't build a self-driving car company without advanced tech.

After my time at Google working on adaptive learning, I started collaborating with Caesar, Felix, who became my co-founders, along with a handful of other former Google coworkers. We were all passionate about FinTech and what technology could do in that space.

We realized that you could pretty much do everything on your phone, except manage your finances in a holistic way. This insight was born out of our own financial planning journeys. We felt strongly that it was time someone built a solution for this, and we wanted to be our own product's users, solving the problems we'd encountered ourselves.

We started thinking about who this product would be for, initially focusing on tech employees. We noticed that unless you have $25 million or more in assets, you pretty much have to manage your own financial journey. If you're in that range of having 500K to 25 million in assets, you're kind of stuck in the middle. You don't have access to the high-quality services that ultra-high net worth individuals get from private banks or family offices, but you're also beyond the basic services offered to those just starting out.

So we thought, how do you serve these users who are somewhere in the middle of their financial journey? These are often tech employees or other affluent individuals who started making decent money early in their careers but aren't quite ultra-high net worth yet. They need more sophisticated financial management tools and services, but they're underserved by the current market.

That's where Arta was born - with a desire to serve these users, to provide them with the kind of comprehensive financial management tools and services that are typically only available to the ultra-wealthy. We're trying to democratize access to high-quality financial management, leveraging technology to make it more accessible and efficient.

Who’s your typical client?

Our typical user is someone who's in that middle ground of wealth management - they're beyond the basics but not quite at the ultra-high net worth level. We're talking about people who have assets ranging from $500K to maybe 25 million dollars. Often, these are tech employees or other affluent individuals who've started making good money early in their careers.

In terms of what we offer, we've got a range of strategies. On the simpler end, we have robo-advisors. We think of these as our 'Model T' - basic, but they get the job done. We take it a step further with deeper insights and more customizations. But we also enable our clients to utilize more sophisticated offerings like quant strategies and direct indexing that could work much better for them in terms of performance or taxes.

With direct indexing, instead of just buying an ETF, we're buying the individual stocks that make up that index. This allows us to do things like tax-loss harvesting as the strategy runs. We also have quant strategies that allow our members to customize their portfolio. For example, if someone's overexposed to tech, they can lean away from that sector. We call this 'defensive growth' - it's about minimizing drawdowns while still trying to keep up with market-like returns.

The goal isn't to chase alpha, but to create a more stable portfolio. This approach worked well for us during the 2022 market downturn caused by rate hikes. And that's important because for a typical tech employee, a lot of their wealth is tied up in tech. When that part of their portfolio is going down, you want other parts to hold steady.

So in essence, we're trying to offer sophisticated wealth management tools and strategies to a group that's been traditionally underserved by the financial industry. And we're using technology to do it in a way that's more accessible and aligned with our users' interests.

What’s your business model? 

As for our business model, we've got a few different approaches. We have strategies that use the standard AUM-based pricing that you're probably familiar with. But we also offer performance-based pricing for some of our strategies, which aligns our interests more closely with our members - we win when they win. However, there are regulatory restrictions on who can access this type of pricing. It's limited to Accredited Investors and above, so we can't offer it to everyone.

How are LLMs going to impact finance?

Financial services are going to experience a big change over the next decade. That's my hypothesis. The reason is, when you look at what LLMs do, they're really good at taking in information, summarizing it, and extracting key points. And if you look at the value chain in finance, a lot of it is just reviewing documents, right? It's all about processing information.

What LLMs allow us to do is drastically simplify these chains of events that happen in financial processes. Take an ACAT transfer as an example, for those of you who are not familiar, this is the system that allows stocks to be transferred from one account to another across brokerages, etc. Right now, you fill out a form, sometimes you even have to fax it, then someone manually enters that data. In the operations process, there are probably five different checkpoints where people are manually checking stuff. It's inefficient and time-consuming.

Now, imagine automating this whole process with LLMs. It's not going to happen overnight, of course. It'll take many iterations. Maybe instead of three manual steps, we get it down to two, with one step handled by LLMs. But you can see how this could lead to experiences that are way more streamlined and real-time. You don't have to wait on people to get your information anymore.

We're already seeing the impact of this kind of technology. For our recent global launch in Singapore, we introduced a feature that generates a live portfolio recap for you every day. It tells you how your portfolio performed yesterday, over the last week, and over the last month. And here's the cool part - you can choose the level of expertise for the report. If you want a detailed analysis like you're a Wall Street analyst, we can do that. If you want something more basic because you're working in tech or you're a doctor, we can do that too.

What’s been surprising to you? 

I think just the pace, the pace at which this technology is moving. It's not surprising that it's getting better - that's expected. But what's really surprising is how fast it's getting better. That cycle is quicker than I expected. 

ICYMI
Recent Five Minutes w/ Interviews

• Finster’s Sid Jayakumar on building AI agents for Wall Street

• Bain’s Richard Lichtenstein on AI adoption in private equity

• Snowflake’s Jonathan Regenstein on AI building novel datasets

• Skadden’s Dan Michael on the SEC’s AI stance

• Stardog’s Matt Lucas on hallucination-Free AI

Thanks for reading!

Drop me a line if you have story ideas, research, or upcoming conferences to share. [email protected]