Opetek has launched ARIUS, an AI reasoning platform for institutional trading desks. The system is being piloted by a Tier One bank in live FX options workflows.
The pilot is running across trading teams in New York, London and Singapore. Traders and quantitative analysts are using the platform to analyse market questions, test scenarios and support live trading decisions.
Opetek is also preparing a further deployment with another Tier One institution, signalling early interest from large banks in tools that work alongside existing trading systems rather than replace them.
Institutional trading desks face growing pressure to make faster decisions while handling rising volumes of data across disconnected systems. Market data, analytics, risk systems, research tools and internal communications often sit in separate environments, slowing analysis and making it harder to trace how a decision was reached.
Against that backdrop, ARIUS was developed as a front-office tool that links market data, pricing models, analytics and news in a single workflow. The platform produces outputs users can inspect, including calculations, assumptions, methodologies, code and market data inputs.
That audit trail addresses a key concern for banks adopting AI in sensitive trading functions. In capital markets, firms must be able to review and challenge decision-support tools, particularly when outputs may influence trades, hedging activity or pricing decisions.
How it works
ARIUS uses a natural-language interface that allows traders and quants to generate bespoke analysis and code more quickly. Delivered as a standalone application that sits above existing data, analytics and model layers, it lets banks test the platform without replacing core trading infrastructure.
ARIUS currently integrates with Refinitiv/LSEG market data and Opetek's own pricing and analytics tools. FX options is the first asset class targeted by the system, although the same underlying architecture is intended for equities, fixed income, commodities, credit and digital assets.
The launch comes as financial firms look beyond general-purpose AI assistants and focus on systems tailored to tightly controlled workflows. In trading, that has increased demand for tools that can show how an output was generated and let users inspect the chain of reasoning.
Opetek was founded by Varqa Abyaneh, whose previous roles include Chief Product Officer at Quantile, the post-trade optimisation business acquired by LSEG. His background also includes senior positions in derivatives trading, quantitative analysis and product development at banks including HSBC.
The wider founding team includes Philip Beadling, Chief Technology Officer and former Chief Architect at Quantile, and Henry Ip, Machine Learning Lead. The mix of trading, quantitative and engineering experience reflects the company's focus on front-office use cases in capital markets.
Decision pressure
Trading desks have long relied on a mix of pricing models, spreadsheets, market terminals, internal chat systems and one-off analysis written in code. That patchwork can create delays when traders need to compare data sources, test scenarios or challenge assumptions during fast-moving markets.
Manual work adds further strain. Bespoke analysis often requires time from quantitative teams or individual traders, while undocumented assumptions can make it difficult for firms to reproduce decisions later or assess whether processes were followed consistently.
For banks, the appeal of reasoning-based systems lies partly in their ability to reduce that friction without creating a new black box. Tools used in markets are subject not only to commercial scrutiny from desk heads and risk managers but also to governance requirements around reviewability and control.
Abyaneh said the company built ARIUS with that tension in mind. "One of the most fundamental challenges in financial markets is deciding what to trade. Reaching that decision requires quantitative analysis and qualitative judgement across market data, pricing models, analytics and news, often under severe time pressure. Because these are high-value decisions, speed alone is not enough. Trading desks need reasoning that is explainable, auditable, and reproducible," he said.
He added: "We built ARIUS around that complexity. ARIUS helps trading desks reason better and decide faster, while keeping the reasoning visible and under human control."