Ultramarin

The removal test leaves nothing at all. The platform is vertically integrated machine learning running from data acquisition through prediction models to portfolio optimisation and execution, described by the company as an operating system for fully automated investment processes, and built since 2017 by a team spanning machine learning, mathematics, physics and neuroscience alongside asset management. Its stated purpose is identifying non linear relationships that conventional econometric models miss, which is a claim only models can make.

The company describes its platform as an operating system for fully automated investment processes spanning data, prediction, portfolio optimisation and execution, which places automation across the entire chain including the point where orders reach the market. No human checkpoint, review step, override mechanism or intervention threshold is described anywhere.

Customisation exists at the level of constraints rather than of supervision, with clients setting exposure limits and risk parameters that the system then operates within, which is governance by boundary rather than by oversight. For a supervised manager that arrangement will exist internally, and none of it is published.

Explainability is the stated differentiator of the research interface rather than a general claim, promising transparency in artificial intelligence driven investment decisions, and the company describes full transparency in decision making as a property of its tailored solutions.

The architecture itself is disclosed with unusual specificity, including a dual process approach to tactical allocation drawn explicitly from behavioural economics that models intuitive and deliberative market behaviour as an early warning system, which is a stated design rationale rather than an unexplained black box.

The strongest validation is structural: the company runs over a billion euros of assets on this platform under regulatory supervision, which is skin in the game no marketing claim matches. No performance, backtest or attribution figures are published.

The evidence is that the platform runs real money at scale under supervision. A licensed asset management subsidiary crossed one billion euros of assets under management for the first time in December 2025, having held a national regulator's licence since September 2024, and the company describes that business explicitly as proving the platform at scale.

Products are live across active exchange traded funds, mutual funds, managed accounts and individual institutional mandates, and the research platform covers more than 2,000 companies globally. A ten million euro Series A was led by an established European growth investor with a prominent former corporate chief financial officer among the shareholders, and a recognised figure in quantitative asset management joined the leadership in 2020. Three German offices support it.

No boundary statement was located, and this company's structure raises a conflict question sharper than the usual pooling concern. It manages more than a billion euros of its own funds on the same platform it licenses to banks and intermediaries, so a client integrating the research interface is consuming signals from a manager that trades on them, and nothing states whether both receive the same output at the same time or how sequencing is handled. The reverse also applies, since queries through the interface reveal what an institutional client is researching. Neither direction is addressed.

Structurally favourable because the payload is market data, company fundamentals and behavioural market signals rather than information about people, so no consumer is the subject of any assessment and the usual concerns of this axis do not arise. What is confidential is institutional rather than personal, namely a client's mandate parameters, exposure constraints and holdings. Held at B because no data processing terms, retention schedule or subprocessor list was located.

No attestation, certification, trust centre or enumerated framework was located. Holding a financial regulator's licence brings operational resilience and outsourcing requirements that the supervised entity must satisfy, so controls exist and are examined, and none of that is published in a form a prospective interface client could assess.

This company is directly supervised rather than adjacent to supervision. Its asset management subsidiary has held a national financial regulator's licence since September 2024 and operates under it across active exchange traded funds and individual institutional mandates, which subjects the investment process, its governance and its disclosures to examination.

European sustainability disclosure obligations are named at the level of specific articles, with strategies offered under either the basic or the promotional sustainability regime, alongside sustainability tilting and client specific preferences. Few vendors in this index hold a licence at all, and fewer still name the disclosure framework their products sit under.

The founder makes the inverse of the usual argument, contending that the opportunity to reduce human bias in investment decisions is vast and that heterogeneous data combined with holistic valuation models produces better strategies, which is fair as far as it goes since behavioural bias in discretionary investing is among the best documented effects in finance.

What is unaddressed is that model bias substitutes for human bias rather than eliminating it, and nothing describes testing for systematic tilts, factor concentration or crowding across the strategies the platform generates. Sustainability tilting adds a further dimension, since the models encode judgements about which companies deserve capital, and no methodology for those judgements is published.

No guarantee, indemnity or falsifiable commitment was located for the technology product. Investors in the managed vehicles hold the recourse that fund regulation provides, including disclosure duties, supervision of the manager and the ordinary remedies of investment law, which is meaningful but belongs to the licensed business rather than to the platform. A bank integrating the research interface has nothing described: no statement of what is owed if a signal is wrong, no correction process, and no service commitment.

Input categories are described rather than sourced, spanning heterogeneous data sets including fundamental analysis and behavioural finance inputs across global equity markets, with the models themselves built in house since 2017. For a research product covering more than 2,000 companies, the identity, licensing and refresh characteristics of the underlying market and fundamental data determine both coverage and cost, and no provider is named. No hosting or subprocessor arrangement appears.

Five consumption routes are offered and they cover the practical constraints institutional buyers face, since a bank that cannot integrate a research interface can buy an exchange traded fund, and one that wants signals inside its own process can take direct interface access to research on more than 2,000 companies. That flexibility is itself the integration strategy. What is absent is any named platform, data provider, custodian or order management system, and no developer documentation was located.

No hosting provider, region selection, residency commitment or private deployment option was located. Operating from three German offices under national supervision implies European processing without stating it, and institutional clients integrating the research interface into their own investment processes would expect the arrangement documented.

No pricing is published for the research interface and no fee schedule appears for the managed vehicles, though fund vehicles necessarily disclose charges in their own documentation. What is unusually clear is the range of commercial forms available, with clients able to take the same intelligence as a fund, a managed account, an exchange traded product, a structured product or a direct interface integration, which lets a buyer choose the wrapper that suits their mandate. The cost of each is undescribed.

Buyers span banks, family offices, financial intermediaries and institutional investors, and the delivery options are broad enough that each can consume the platform in the form their mandate permits. Subject coverage is global equity markets with research on more than 2,000 companies, and customisation is genuinely deep, with exposure controls configurable by sector, currency, region, total market and specific factors, plus sustainability preferences. The limit is asset class: this is equities, with no fixed income, credit or multi asset capability described.