arbitAI, a new Hong Kong firm specialising in deep learning technologies and neural networks, has hired Oliviu Chivu as chief technology officer.
Prior to joining the tech startup, Chivu worked with Russia’s largest bank Sberbank, where he collaborated with GridGain, a US consultancy that offeres in-memory computing services for big-data systems.
With Chivu’s help, Sberbank was an early adopter of GridGain’s new database technology, enabling it to process an almost unimaginable volume of a claimed one billion transactions per second on computer hardware said to have cost only US$25,000.
At arbitAI, claims Chivu, “the use of a RAM-based in-memory computing solution will decrease costs while drastically improving performance and scalability”. His vast experience, says Ionut Sarbulescu, co-founder and chief executive officer of arbitAI, “is a crucial advantage for us in reducing latency, thereby improving the speed and efficiency of our algorithms”.
Before Sberbank, Chivu worked in a similar role in the Singapore office of Virtu Financial, a leading electronic market-maker and HFT firm head-quartered in New York.
HFT firms like Virtu Financial employ sophisticated machine-learning algorithms to seek differences between the various trading prices quoted for listed securities among different market participants. Then, HFT firms buy and sell those securities in a matter of fractions of seconds, securing a small profit.
Since the price differences are typically very small and exist only briefly, traders need to act upon them within milliseconds. That time difference between the discovery of a pricing anomaly and the execution of a trade is known as latency. HFT practitioners go to great lengths – and, often, spend millions of dollars – to reduce that difference and, thereby, increase their profits.
Chivu’s main achievement at Virtu Financial was the implementation, and then further development, of the firm’s low-latency trading infrastructure.