Benzinga Recap: How Devexperts Manages Data and AI with Vitaly Kudinov
Vitaly Kudinov, SVP of Sales and Business Development at Devexperts, shared insights during a recent panel session at Benzinga’s Future of Digital Assets in New York.
Vitaly explored topics like how Devexperts handles data and integrates AI into its systems and products, and future developments.
This article provides a recap of his talking points, which touch on trending issues in the brokerage industry.
Collecting, storing, and using data effectively
Vitaly began by making a distinction between private data and public data in the brokerage space.
Private data
Private data is that which comes from the broker’s own platform. Devexperts only builds tools to help brokers analyze and work with their private data.
“This is the broker’s data—client information, trades, and the like. It’s their data, and we never share it with anyone else,” Vitaly explained.
Public data
This refers to the data that is public by default, such as market data and news. Vitaly revealed the scale of Devexperts’ public data repository:
“There’s a storage of 5–6 petabytes of information. We’ve been collecting every tick since around 2010, and for daily aggregations, we go back to 1970.”
Data lakes
Devexperts recognized that traditional data extraction methods were inefficient, and developed data lakes to streamline access.
“At some point, we realized that this data—even if we package it well—takes too long to extract, so we started building data lakes on top of the original storage.”
These data lakes make historical data available in a useful format for market players, such as quants and buy-side and sell-side firms.
| 💡What are data lakes? Data lakes are centralized repositories for storing large amounts of raw, unstructured, and structured data at any scale. Unlike traditional databases, which require data to be organized and processed before storage, data lakes allow companies to store data in its original form and process it later as needed. |
API layer
Next up, Devexperts built an API layer that can be hosted on its premises, or the customer’s premises (in a colocation facility or data center for example).
“We improved the transport layer to the stage where it doesn’t matter if you’re using frontend technology like a web browser, mobile app, or backend technology, you are using the same API and you get the fastest access to historical data,” highlighted Vitaly.
| 💡What’s an API layer? An API layer acts as a bridge between different software systems, allowing them to communicate and share data. It simplifies the process by allowing developers to access tools or data without dealing with the complex details behind how everything works. In fintech, an API layer is crucial for integrating platforms like trading systems, mobile apps, and data analytics tools. In a nutshell—API layers make sure data exchange across devices and applications is fast, secure, and efficient. |
AI & data at Devexperts
“It changed the rules of the game,” said Vitaly as he entered the topic of AI and data. He went on to explain the transition between queries for structured data and the intuitive, natural-language interactions that AI has made possible.
“Before AI, we were working with structured data, which is like SQL databases or other types of storage, where you had to create a query to get the data. Whereas, AI doesn’t require you to formulate a query. Instead, you speak in natural language, and the AI provides you with the answer.”
Devexperts’ background with AI
Vitaly explained that Devexperts had actually been working on AI long before large language models (LLMs) became mainstream.
“We started with low-hanging fruits. Not with LLMs even, because AI is in fact broader than just LLMs and ChatGPT. We actually started working on this AI project seven years ago, when there was no LLM available in the market.”
| 💡What are LLMs? Large Language Models are advanced AI systems trained on large amounts of text data to understand and generate human-like language. They can perform tasks like answering questions, translating languages, creating content, and summarizing information. ChatGPT is a familiar example of an LLM. |
Devexperts is now converting the huge amount of data that it has, into a format that AI can effectively consume.
AI powered by data directly from the trading platform
Devexa is Devexperts’ AI-powered chatbot solution. Vitaly leaned on the example of Devexa to illustrate how data directly from the trading platform can be used to provide trustworthy and reliable AI-powered support.
“Because the widget is connected to the trading platform, if the trader allows, it can use data such as the trader’s account, positions, cash balances, and so on to provide support.”
For context, Devexa can be accessed by end-users as a widget inside a trading platform, as a widget inside a website, or even directly via their favorite messaging app— like WhatsApp, Telegram, or Discord, for example.
Avoiding AI hallucinations and ensuring valuable support
When asked how Devexperts makes sure Devexa does not hallucinate and provides trustworthy, accurate, and reliable support, Vitaly discussed three main points:
- Improving communication between the end client and the broker
Devexa is connected directly to the trading platform, answering questions that would typically be handled by a human agent.
The difference is that with Devexa users don’t need to wait for the agent to take their call or respond to their live chat. Devexa can instantly extract the account information.
“There is no place for hallucination because we are not training this AI on external data, we are providing the data in the best format for the user, and we provide this immediately.”
- Multi-language capabilities
Auto-translation capabilities solve the challenge of hiring customer service agents who speak a wide range of languages.
For example, traders can send a voice note in their native language, and Devexa will respond with accurate text in their language, complete with an understanding of trader nuances and lingo.
- Balancing Devexa with human support
Vitaly made clear that a hybrid approach of both AI and human support is necessary. He explained that although Devexa is a huge win in terms of optimizing personnel resources for brokers, human support is still required to strike the right balance.
“What you can automate, you automate—KYC, onboarding—that’s doable, very easy, but you can’t allow AI to have full control.”
More complicated queries are still routed to agents, however, they too are able to use Devexa to provide AI support. For example, they can screen share and video call with the client using Devexa’s admin panel.
“And all of these functions didn’t even require LLMs”, added Vitaly, illustrating Devexpert’s approach to AI support further.”
Any exciting developments on the cards?
When asked about what he was most excited about at Devexperts, Vitaly shared about their new <web widget library>.
“Our dev team can now build production-ready widgets in just a couple of days. Using the same API layer, these widgets are cost-effective and highly customizable. It took a long time to get here—but now that we’re here, it’s pretty incredible.”
This new offering will mean brokers can quickly adapt to new market demands and introduce new features and functionality without significant development.
Connect with Vitaly and Devexperts at future events
Devexperts participates in a large number of fintech events like Benzinga each year. At each event, Devexperts has a team on hand ready to share their expertise. Head over to Devexperts events page for more information.
Connect with Vitaly on LinkedIn.
Connect with Devexperts on LinkedIn.