Squared-S, a Hong Kong-based software startup, is set to sign an agreement with Mirae Asset Global and Hong Kong University of Science and Technology’s business school to develop asset-allocation tools driven by artificial intelligence.
“This is a research collaboration,” said Soujit Ghosh, co-founder and CEO at Squared-S, to build on pending patents held by the company that use A.I. to optimize investment portfolio performance. “We’re jointly building to develop A.I.-driven intellectual property to be used in Mirae’s global fund efforts.”
A Mirae spokeswoman declined to comment, noting the agreement has not yet been finalized.
For now the fintech’s emphasis is on using A.I. to help portfolio managers control volatility, correlations, leverage and other risk factors, rather than trying to identify alpha-generating securities or predict market directions.
Modeling the universe
“Most portfolio optmizers just throw allocation numbers or percentages at a portfolio,” Ghosh said. “We enhance that process. Fund managers can still do research and pick stocks, and our software adjusts to their mandate.”
Seth Huang, co-founder and chief A.I. officer at Squared-S, likens the company’s software to AlphaGo Zero, the second version of Deep Mind’s AlphaGo A.I., which gamed fame for defeating human masters of the game of go. Whereas AlphaGo achieved this based on combining a series of algorithms, AlphaGo Zero consolidates these into one system. This lets its A.I. make more decisions within the model.
“The methodology is open-sourced,” Huang said. “The hard part is designing it to apply to financial markets. Historical data doesn’t get repeated. Life isn’t a game of go.”
Squared-S’s models attempt to map the entire financial universe. “All products globally are interconnected,” Huang said. “Capital flows are interconnected, like undercurrents in the ocean. The A.I. is meant to model the structure of the universe and how the universe evolves.”
From funds to structured products
That ambitious description suggests that Squared-S’s ambitions extend beyond risk-managing portfolios.
Down the road, the founders want to address structured products, such as warrants in Hong Kong and accumulators in Korea. The A.I. can generate ideas for product designs, from underlyings to derivative structures.
It can also apply to making trading, rebalancing decisions and hedging strategies more efficient, ultimately putting A.I. at the center of how global banks manage their derivatives positions.
“Structured products are a $100 billion business and the hardest part of trading at an institutional desk,” said Ghosh.
But the founders are aware that many people in the finance industry view A.I. as job-killers – which they deny, although they acknowledge it can change the nature of investment, analyst and trader work.
That’s another reason why they want to debut with risk tools for traditional fund managers. “Being a bottom-up stockpicker doesn’t meant A.I. can’t help you,” Huang said.
“We’ve got to treat the industry with a bit of respect, and show them the benefits of A.I. step by step,” Ghosh added.
Ghosh comes from the industry, as a former equities derivatives trader at Goldman Sachs and J.P. Morgan. Huang is an academic who has also worked for equity long/short funds around the region. The third partner is Rachid Ait Seddik, the company’s machine-learning system architect.
Although Ghosh and Huang met in Seoul in 2014, their lives took them separately to Hong Kong, where they reunited with a sense that banks were ignoring opportunities to automate derivatives trading and structured products.
They are currently seeking seed funding. “We’re in conversations with venture capital,” Ghosh said, noting a deal with Mirae should make a big difference to investors.
The funding is needed to hire more software specialists who can deliver A.I. solutions to clients, including data visualization tools. “The engine is there; we just need manpower,” he said.