Beeks Financial Cloud Group has partnered with Canadian fintech OmniPhi to develop a sovereign agentic AI platform for capital markets, combining privately hosted large language models with tools for automating trading-related workflows. The agreement includes a strategic investment by Beeks in OmniPhi and will see the companies target banks, brokers, asset managers and professional traders.
Beeks Sovereign AI is designed to overcome one of the barriers to institutional adoption of agentic AI: the risk of exposing sensitive information to external AI providers. Using third-party platforms for tasks such as research, strategy development and code generation can involve sending proprietary trading data, algorithms and other confidential information outside an institution’s own environment. The Beeks/OmniPhi approach is intended to keep sensitive information within infrastructure controlled by the institution.Three Tiers of AI Infrastructure
Beeks plans to offer the platform through three service tiers. The entry-level offering will provide shared GPU resources and AI models through virtual machine-based infrastructure, primarily targeting sophisticated retail and professional traders. A mid-market service will use GPU-enabled versions of Beeks’ existing Proximity Cloud infrastructure, allowing customers to access dedicated computing resources.
“At the higher end, we’ll offer fully dedicated cabinets of GPUs with the ability to run models locally,” explains Gordon McArthur, CEO of Beeks, in conversation with TradingTech Insight. “We expect Tier 1 institutions to have their own sovereign AI environments, which can run either in colocation facilities or deployed in their own data centres. The data never needs to leave the building, and we think that’s a particularly strong selling point for large institutions.”
The companies are building infrastructure in New York and London, with the full dedicated offering targeted for the first quarter of 2027. OmniPhi’s platform is currently in beta, with test environments already available and around 7,000 retail users on its waiting list.
Dedicated deployments could also allow institutions to move away from token-based charges towards capacity-based arrangements, with GPU requirements determined by the models and workloads they intend to run.
From Trading Ideas to Production
“The core engine is Phi, our agentic harness,” says Pouya Farmand, Co-Founder and President of OmniPhi. “For institutions, one of the key capabilities is that we can build bespoke tools and connect the platform to their existing internal systems, as well as external tools and services. You can connect it to data sources, brokers, Bloomberg and other systems. The environment then adapts to the user’s intentions and trading approach over time, so a scalper and a portfolio manager will end up with very different environments.”
For quantitative investment firms, the technology could reduce the time and resources required to move from an initial trading idea to a tested, deployable strategy. McArthur describes how he has been evaluating the platform through his own algorithmic trading operation, providing Beeks with first-hand experience of its capabilities.
“I’ve personally run a small algorithmic trading operation for about 20 years. Historically, it could take me nine months to go from designing an algorithm to getting it into production. I’ve developed five in two months using OmniPhi. I gave it the criteria for trading silver futures, and it came back with five strategies, each with a two-year backtest, walk-forward analysis and Monte Carlo analysis. It shortened my development cycle from months to days.”
Sell-side applications remain less clearly defined. Beeks is exploring proof-of-concept opportunities with banks and brokers, including potential uses in order flow optimisation and accelerating the development of trading applications. The companies are particularly interested in how the technology could support existing quant and development teams while operating alongside established low-latency execution infrastructure.
Expanding into Institutional Markets
OmniPhi had been self-financed by its founders and had begun generating revenue before Beeks’ investment. The partnership also gives Beeks an opportunity to extend its existing institutional infrastructure business into a new category of AI services.
“The partnership is particularly important because it allows us to scale much faster,” says Ali Mokhtari, Co-Founder and CEO of OmniPhi. “We already have a substantial waiting list, but we needed to build the support capabilities, tools and infrastructure to serve those users, continue improving the platform and expand the engineering team. Beeks’ investment helps us do that. It also supports our marketing efforts for our core product and opens up a much bigger opportunity in the institutional market, which gives us an entirely new perspective on how the technology can be used.”
Both companies intend to maintain a focus on professional retail traders alongside their institutional ambitions. McArthur acknowledges that effective use of the technology still requires financial markets expertise, particularly the ability to formulate appropriate instructions and evaluate the resulting strategies.
With dedicated infrastructure deployments planned for early 2027, the companies are seeking institutional proof-of-concept projects that could help establish where agentic AI delivers the greatest operational and commercial value across capital markets.
Subscribe to our newsletter
