Pavel Perfilov is a professional working at the intersection of modern technology and finance, with extensive experience in low latency trading, market data management, and pre- and post-trade risk management. Pavel is particularly knowledgeable about the complexities of brokerage and exchange infrastructure due to his extensive experience with various exchanges, which includes everything from paperwork to patching and physically installing servers. In this talk, he discusses technology trends in trading and the future of finance in the aftermath of a new wave of technological revolution that has brought AI into our daily lives.
As an ex-IT director of an electronic brokerage company, can you tell us a bit about the tools and methodologies for performance testing and monitoring of trading systems that you find the most effective at the moment?
Electronic trading services (DMA) are a very demanding sector; clients require ultra-fast speed of processing orders, higher leverage, better location, and attractive fees. All of these add complexity to the process of building robust and reliable trading systems. You can’t simply add extra checks and additional logging because it would add delays, queues, and non-deterministic behavior, which clients do not usually like.
One of the solutions for building high-performance systems for monitoring is to build post-trade monitoring systems and systems that capture traffic and process all messages that are being transmitted over the network. This allows it to minimize its impact on the trading system itself and allows it to monitor every bit of information transmitted. This approach doesn’t work very well in risk systems, as there are some extra regulatory requirements that are hard to solve in a way that won’t impact speed. One of the examples is the requirement to record client transactions and the state of his account before and after the transaction. These functional parts, which are making fast pre- and post-trade checks and record keeping, are typically treated as ‘secret sauce’ in organizations. With regards to methodologies and tooling, modern CI/CD pipelines and automated testing, which include performance testing with very high loads, work very well. Sometimes replaying traffic is considered a good tool for finding bugs and issues before production rollout.
How are evolving technologies like artificial intelligence and machine learning transforming the trading sector?
The brokerage and finance industry is very regulated, so sell-side is simply not allowed to build AI and non-linear neural networks in clients’ order processing. At the same time, in non-regulated areas where the client takes responsibility for their decisions, AI has become a game changer. For example, some brokers give extra tooling to a client to compose a portfolio in the most efficient manner. AI allows them to extract sentiments from the news feeds and label some market moves by the events that happened at the time so that the clients can better understand and learn market behavior patterns. I think we’re just at the beginning of a new era of AI-driven financial products.
What challenges and opportunities do you see in the integration of AI?
Challenges are not visible yet because the banking sector is about trust and reliability. Unicorn startups are not changing the game (yet) because of a lack of trust, so big clients tend to stay with companies that have a proven track record and a long history. Modern companies attract more young clients because they offer gamified products and services. If unicorn companies managed to be successful in the next 10 years, the structure of the brokerage industry would probably change.
What’s your take on the future of decentralized finance?
The ideas and concepts behind DeFi are very good. I see that performance and capacity, first of all, have significantly improved, so tens of millions of transactions are not blocked anymore. This made me think that the industry is growing and that it’s great to have more venues where clients can find liquidity. I don’t think it’s a threat to the real financial sector.
What future applications do you foresee for blockchain technology in financial markets?
First of all, I think most people have realized that this technology is real and that it works. Crypto exchanges and De-Fi are operating, volumes traded are growing, and regulation is also changing, so we are slowly transforming the paradigm of classical financial services. The future is unknown, but as technology proves more and more reliable, low-cost, and efficient in operations, I think it’s inevitable that it will get more credit. I think trust here is the main enabler; the more trust a certain technology has, the more chances it would have to get approved by regulators. If regulators approve it, there will be a rapid growth of the services, but I hope it will be controlled and the quality of the new products won’t impact reliability and trustworthiness.
What strategies do you employ to continuously innovate and stay ahead in the competitive fintech landscape?
As of now, it’s all about cost and speed. The strategy is to minimize transactional and operational costs. Fee-free brokerage services are sort of a new reality that currently drives the market. This trend pushes brokers to find new areas for service monetization. The strategy is simple here: always try something new, be it AI, ML, blockchain, or something else. If you find a way to build an innovation pipeline in an efficient manner that doesn’t have a high cost or time to market, you’ll be able to collect feedback and get some sense of the potential value of the innovations, which could lead to higher returns.
How do you envision the finance industry evolving over the next five years? What current practices will become obsolete, and what new trends do you anticipate emerging?
Five years ago, we were not able to imagine 0-fee brokers, AI-driven automated advisory services, and such an amount of ML pipelines in every department and every service. The cost structure has changed significantly, the velocity of the market has changed, and the trend here is faster time-to-market for new smart finance products. Speed and fees are at their physical limits, so it’s not attractive anymore. So the industry would go towards some ‘smart products’ that are smoothly integrated into our daily lives.
Featured image: Unsplash
