Trade and regulatory reporting specialist Xinthesys has launched AIMOS, an agentic artificial intelligence framework within its Agile Data Processing and Engineering Technology Platform (ADEPT) and the underlying event-based TradeStore repository. TradeStore is designed to be a single source of truth for event-level trade lifecycle data at scale, supporting sell-side firms processing hundreds of millions of transactions per day.
AIMOS builds on that infrastructure, adding a natural-language interface intended to help users interrogate retained trade records and automate workflows around those records.
Reporting Origins
Founder and chief executive Jeffrey Stein described the origins of the ADEPT platform in conversation with RegTech Insight. During an earlier consulting assignment with a large swaps dealer, he described an environment where each asset class was supported by four sets of trading systems. Assembling the records needed to answer a regulatory enquiry could take weeks or months.That experience laid the conceptual foundations for TradeStore, a platform that could capture trade lifecycle activity as an event log and preserve its relationship with derived lifecycle data and reporting outputs. TradeStore has been operational for about seven years.
Reporting coverage includes CAT/CAIS Submissions, Rule 605/606, TRACE & TRF, LOPR. Controls include Evidence & Lineage and Surveillance.
Event Lineage
TradeStore groups lifecycle events into what Stein describes as a “logical trade”. Each event has a human-readable representation linked to the raw message received from the source system. Developers use the ADEPT platform to build code that processes those events at scale, with outputs linked back to their originating events in TradeStore. AIMOS takes a natural language request and automates the workflow and data access code generation.
Stein illustrated the structure with a cancellation. An investigator can examine the cancellation event, inspect its corresponding CAT report and trace the record back to the original message. The repository retains the source attributes alongside the event and reporting output. “It’s exactly the facts of the trade: what was reported, what was captured and what attributes came in on the line when the trade came in,” he says.
ADEPT supports workflows built around data pipelines that create trade lifecycle events and derived data, including regulatory reports. A key foundational component is the ADEPT TradeStore that supplies the event records and has pre-built services that allow development teams to quickly build data processing workflows using assisted AI development.
AIMOS is designed to respond to a natural language regulatory request such as: “Can you pull me all of the orders reported to CAT for this date range and validate the report against the underlying events and detail level”.
“Unlike traditional AI assistants that primarily generate or summarize information, AIMOS is designed to understand an objective, identify the data and actions required, execute a workflow and produce a structured, usable result,” Stein said in the launch announcement.
The current release of AIMOS is designed to respond to a natural language regulatory request such as: “Can you pull me all of the orders reported to CAT for this date range and validate the report against the underlying events and detail level”.
Workflow Automation
AIMOS is designed to be an extensible architecture and among future use-cases Stein describes two trade surveillance challenges: adapting detection rules to different firms’ trading activities and running those rules across high volumes of orders. AIMOS is intended to let users describe a detection requirement or modify an existing rule, then generate the code for ADEPT to execute. For example, a trading desk could adapt a wash-trading rule to the attributes in its trade messages without writing any code. The planned approach would use AI to construct the detection logic, while ADEPT executes that code at the speed required to process trading activity at scale.
Xinthesys is focusing on mid-tier broker-dealers seeking a way to understand how AI can transform operational efficiency AIMOS is intended to give those firms a low-impact, low-cost way to trial the investigation of retained records and develop further workflows without the need to commission each task through a technology team.
Subscribe to our newsletter
