Machine Learning Hedge Fund Hires AWS’s Lead Scientist and Point72 Veteran Andrew Arnold – Institutional Investor

Delphia, an artificial intelligence-based investment advisor, created a buzz when it decided to build a proprietary data strategy that compensates individual investors who share their personal data for a chance to improve their returns. Now the investment advisor — which is part app, part hedge fund — is creating a stir by hiring Amazon Web Services’ lead scientist, Andrew Arnold. Arnold is an expert in AI who worked for Steve Cohen’s Cubist Systematic Strategies (Point72) for three years before AWS. As Delphia’s chief scientist, Arnold will help the firm push deeper into AI and machine-learning tools and modeling. Delphia offers actively managed long-only strategies for individuals and an equity market neutral hedge fund for accredited investors. 

Founder and CEO Andrew Peek believes that some of the world’s biggest institutional investors will also contribute data. It’s a bold move. When asked whether Delphia believes that pensions, endowments or others could be convinced to provide their data, he replied, “We do, actually.”

“The Incentive structure that we’ve been developing is inclusive of not just individuals contributing data, but eventually corporates or institutions themselves. We are designing a reward system that handles attribution and acknowledges where it originated from.” Peek, speaking from Delphia’s Toronto headquarters, said the firm still faces all the challenges that an emerging manager without a three-year track record experiences in the institutional market. In addition, the firm will need to assure the world’s biggest investors that their data is safe, and that it will only be used on an aggregate basis. 

Arnold’s hiring comes at a good time for the company. Delphia’s quantitative equity market neutral hedge fund generated 72 percent net returns in its first year. Through the end of June, it has returned 23 percent, on a gross basis, with an annualized Sharpe ratio of three. A Sharpe ratio — which measures the return of an investment relative to its risk — above 3 is considered to be excellent. Although one year is still only one year, it’s also been one of the toughest years in a long time.

Delphia’s quant equity long-short hedge fund strategy uses machine-learning algorithms to make predictions about the fundamentals of 3,000 U.S. equities. Although there are hedge funds that use AI exclusively, Peek said the firm can provide more transparency about the algorithm’s decisions to investors. Delphia uses a framework that is economically intuitive and that dates back 10 years, to a time when his partner, Jonathan Briggs, was managing director of quant equity at CPP Investments. The idea is to use machine learning and AI to skip over the noise created in the short term. Delphia forecasts the fundamentals of thousands of equities for the most immediate quarter and the following six quarters. Although seven quarters is hardly long term, it is when compared to most quant hedge funds, which may turn over positions on a weekly or even daily basis. Delphia’s average holding period is four months.

The next phase for the firm is data, which it will only use to further improve its stock-selection algorithm. “That’s our next chapter,” said Peek. In two years, “Alpha will be mostly derived from that kind of data,” he said. Unlike most hedge funds, Delphia raised $60 million in Series A funding in June, led by Multicoin Capital and with additional participation from, among others, Ribbit Capital, FTX Ventures, Valor Equity Partners, and FJ Labs.

Delphia built a hedge fund that is a systematic version of a human portfolio manager scouring fundamental financial data. With the hiring of Arnold, Delphia will be able to turn a discretionary manager into a “quant investor’s dream,” as Peek described it. Here’s the CEO’s vision: “Discretionary managers also have flaws. Sometimes they are right in the fullness of time, but it takes two years for their bet to come to fruition. That is an inefficient use of the capital,” said Peek. Machine learning can help the firm skip over that short-term noise. “So that’s why you have to think of a larger market structure, not just the skill of an individual.” 

Arnold was the lead scientist on Amazon Web Services’ CodeWhisperer, an AI pair programming tool. At Cubist, Arnold created a hybrid research and trading group focused on machine learning-based signals and strategies. Arnold, who wasn’t available for an interview, said in a statement, “It’s not every day that an institutional-grade investment algorithm makes its way into the hands of everyday people. As markets continue to evolve, data will remain the most essential differentiator needed to stay ahead, and I’m excited to contribute to the incredible amount of innovation already happening within the team.”

Spread the love

Leave a Reply

Your email address will not be published.