Custom AI solutions built on open source
We build AI and data solutions that fit the way you work: on open models, with a thin custom layer, deployed onsite at your company. No per-seat licences and no lock-in to a single vendor.
Platforms built for large corporations come with per-seat licences, long rollouts and the need for in-house specialists. For a company with 30 or 300 employees, the effort is often out of proportion to the benefit.
General-purpose AI assistants know neither your documents nor your workflows. What's left is copying, pasting and fixing – and a tool nobody uses day to day.
Many services process data outside the EU. If you work with customer, staff or machine data, you want to know where it is stored and who can access it.
We pick the right open model for each project – for example Llama 3.1 8B for text, SigLIP for images, or a Whisper model fine-tuned for air traffic control speech. Where a hosted model from a European provider is the better trade-off, we say so.
We build only what your workflow needs: rules, checks, interfaces and a front end your team understands. In our projects, the real value sits in this layer – not in the model itself.
Installation on your hardware or with an EU hosting provider, integration with your systems, documented handover and training for your team. With an on-premise installation, your data never leaves your network.
Five services, one principle: understand first, start small, then deploy onsite.
Data
R and Python pipelines that clean, merge and regularly report on your data – with the rigour of 20 years of clinical-trial statistics: every number traceable, every analysis reproducible.
Documents & AI
AI applications for one clearly defined task – built on open models and backed by transparent rules.
Marketing
Content pipelines that turn raw material into ready-to-post output. Your team keeps editorial control.
Automation
One department, one clear workflow, one measurable result: we automate where manual work costs time today – and extend the open-source tools you already use instead of replacing them.
Deployment
So the solution is actually used: installation, integration, handover and support – on retainer if you wish.
You want to try AI in one area first – admin, accounting or marketing, for example – and only go further once there's a solid result.
Manufacturing with machine and sensor data, service firms with lots of documents, real estate and hospitality with booking and operating data: wherever data piles up but nobody has time to analyse it.
Aircraft owners, clubs and flight schools. Our founder is a private pilot, so we know engine data, radio calls and everyday life at the airfield first-hand.
By phone, video call or in person. You describe the task; we tell you honestly whether AI makes sense for it – and if it doesn't, we say so.
We review your workflows and data and identify the one or two use cases with the best ratio of effort to benefit. The result: a written recommendation.
A working solution on your real data (proof of concept). You decide based on results, not slides.
Installation, integration, training and documentation. After that, maintenance as needed or on retainer.
Each step has a fixed price, which you receive in writing before it starts. You can stop after any step, with no follow-on costs. After the free discovery call you get a short written summary and, if you wish, a quote.
Companies in Bavaria can apply for the Digitalbonus Bayern for digitalisation projects – note: before placing the order. The funding body decides on eligibility; we're happy to help you describe the project.
As of October 2026
No logo wall, no invented references. Here is what we actually built: internal R&D projects, two of them live, and projects for Peony Bloom Estates.
Splits, reads, classifies and names mixed PDF scans – in five languages, free to use.
A 9-page scan with no separator pages: all 4 documents correctly rebuilt.
Turns piston-aircraft engine data into a plain-language debrief – with rule-based checks and cited sources.
Runs on about €5 a month of shared web hosting.
Finds the best scenes in hours of action-cam footage and cuts them into clips and montages.
160 min of footage → 4 highlight clips and 5 montages, with no manual footage review.
An interactive model for a decision on buying two short-term rental units: revenue, costs and break-even – both sides could test their own assumptions.
Live within two weeks, about 1,000 lines of R.
"GDPR-compliant" on its own says little. For every project we state where the models run and where data flows. You choose the setup that fits your data.
| Setup | Where the model runs | Where your data goes | Typical for |
|---|---|---|---|
| On-premise | On your own hardware, in your network | Nowhere – it never leaves your network | Staff, customer and machine data |
| EU hosting | On a server with a hosting provider in the EU | To that provider, covered by a data processing agreement | Applications that must be reachable from outside |
| Hosted model | With the model provider, preferably one based in the EU | The content to be processed goes to the provider – we name it up front | Tasks where a large model is clearly the better choice – as with PaperSweep |
A data processing agreement (Art. 28 GDPR) and our technical and organisational measures are available on request – before the project starts.
For tools that could record employee performance or behaviour, we involve your works council early (§ 87 (1) no. 6 BetrVG).
Article 4 requires measures to support your staff's AI literacy. We train your team in practice – on the tools they actually use.
As of October 2026. Our notes are not legal advice.
AleaNCore is a senior team of one to two people. You talk directly to the person who plans, builds and deploys your system – with no agency mark-up and no hand-off to anyone else.
Arun Lal runs AleaNCore as managing director. He is a statistician and computer scientist and has worked on the analysis and programming of clinical trials since 2006, including for Boehringer Ingelheim, Johnson & Johnson and Grünenthal; in 2015 he founded AleaNCore. That work shaped his standard for data: every number must be traceable and every analysis repeatable.
Today he applies the same approach to AI: open models, transparent rules, documented results. As a private pilot he also knows aviation from the cockpit – which is where HobbsMD and WilcoRoger come from.
Every option has its place. This overview is meant to make the decision easier – even if it isn't us.
| Option | What you get | A good fit when … | How AleaNCore differs |
|---|---|---|---|
| Free public programmes | Information, workshops and orientation, e.g. from Mittelstand-Digital centres or the chamber of commerce (IHK) | You want a first overview – highly recommended | We implement afterwards: after the workshop, we build and deploy the solution. |
| Off-the-shelf AI services | Fast start and polished interfaces, e.g. Microsoft Copilot or ChatGPT Enterprise | The work is general office tasks | Tailored to your data and workflows, your choice of data location, no per-seat licence for your own solution |
| Large consultancies and system integrators | Large teams, certifications, established procurement processes | You're planning a company-wide programme | Direct access to the person doing the work, smaller budgets, short lines of communication |
| Hiring an AI specialist | Full capacity and know-how in-house | AI becomes your core business | No full salary, no long recruitment – you start in weeks, not quarters |
AleaNCore GmbH is an open-source-first AI studio that builds custom AI solutions for SMBs and mid-market companies, deployed onsite by a 1–2 person senior team, at a fraction of enterprise AI cost.
It depends on the scope. Each step – AI check, pilot project, deployment – has a fixed price that you receive in writing before it starts. The first call is free, and you can stop after any step.
Not with an on-premise installation: the model and the data stay in your network. With EU hosting or hosted models, we name the provider, location and data flow up front and sign a data processing agreement.
Not necessarily. Many tasks can be handled by small open models on standard hardware – B-Roll Finder, for example, runs entirely on a MacBook Air. The AI check clarifies what your case needs.
A pilot project usually takes two to four weeks. After that you see a working solution running on your own data.
Every system is handed over with documentation. We build on widely used open tools such as Python and R so that your IT provider can work with them too.
Many typical SMB applications – document processing, reporting, marketing – are not in the high-risk categories. Regardless, Article 4 of the AI Act requires measures to support staff AI literacy; we offer practical training for this. This is not legal advice. (As of October 2026)
Companies in Bavaria can apply for the Digitalbonus Bayern for digitalisation projects – before placing the order. The funding body decides on eligibility. (As of October 2026)
Yes. Onsite work is concentrated between Ulm and Augsburg; visits elsewhere in Germany are arranged on request. Much can also be done by video call.
30 minutes, free and without obligation. Tell us briefly what it's about – we'll reply with suggested times.