FinnVersion and NextEPS: Two Products, One Azure Stack

Datajul builds AI and data systems on Microsoft Azure for clients. We also build our own products, and two of them are now live. This post covers what each one does, the principles behind them, and how they connect to our client work.

FinnVersion: one sentence in, a whole portfolio out

FinnVersion starts from a plain-English investment mandate, such as "profitable mid-caps with low debt and growing free cash flow". It turns that sentence into screening rules, screens the S&P 500 on more than 50 trailing-twelve-month fundamentals, and builds a diversified portfolio.

• Weights are computed in code, never guessed by a model. Position caps and sector limits are applied deterministically, and every cap and redistribution is written to a construction log you can read.
• Every holding has a thesis. You can open any position and read the full investment case, with bear, base and bull valuations and claims cited to filings.
• No look-ahead. Each analysis is pinned to a date, so it only uses information that was available on that date.

Your first portfolio is free, and no card is required.

NextEPS: what will a company earn next quarter?

NextEPS forecasts next-quarter earnings per share for S&P 500 companies. You search a name or ticker, then see past quarters next to the forecast.

• Measured against a simple baseline. A naive forecast assumes a company repeats last quarter's EPS. In a back-test over the last 12 months, NextEPS forecasts landed 19% closer to actual results than that baseline.
• Honest confidence labels. High-confidence forecasts typically land within 12% of the actual figure. Low-confidence forecasts can be off by 50%, and we label them that way.
• Fresh every week. Forecasts refresh every Monday.

Why it all runs on Azure

Both products are built and run on Microsoft Azure. The models are developed with Azure Machine Learning, and the applications are hosted on Azure. That is a deliberate choice. Our client work is Azure work too: private AI assistants on Azure OpenAI, forecasting on Azure Machine Learning, and data platforms on Microsoft Fabric. Running our own products on that stack means we deploy, monitor and pay for the same services we recommend, so we know where they shine and where they need care.

What this means for clients

The habits behind these products carry over to every engagement. We do the math in code where it matters. We back-test models against simple baselines before trusting them. We say clearly how confident a prediction is. And we keep everything inside an Azure environment the owner controls. If you want the same approach for your own data, inside your own tenant, book a free assessment.

FinnVersion and NextEPS are research tools, not investment advice. Datajul LLC is not a registered investment adviser, broker-dealer or financial planner.