01
AI agents & LLM apps
Assistants, agents and automations on your documents, data and processes. Built on the model that fits, from OpenAI, Claude and Gemini to open models on your own servers, and tested before your users see it.
AI, ML and data productsthat work in production.
You know the problem in your business. We build the AI, data or software that solves it, and we stay responsible for it once it's live. Describe what you're working on: get answers, a first sketch and estimate, or a call with us.
This assistant runs on our own hardware in the Netherlands
Any model, your choice: cloud or local.
OpenAI, Claude, Gemini or open models like Llama and Qwen: we pick what fits your data, budget and rules. Open models can also run on your own servers or GPUs, a setup we build on every day: the assistant on this page runs on our own hardware in the Netherlands.
What we do
AI first, on solid foundations.
02
Reliable data and reporting
Your scattered sources brought into one place you and your team can trust: pipelines that run on their own, dashboards that agree with each other, and alerts when something breaks.
03
Machine learning
From problem framing to models in production: forecasting, recommendations, fraud detection, image and text. We start with a baseline and prove the gain with A/B tests.
04
Tools and portals your team can rely on
Internal tools, customer portals and the product around your data or models, including turning an AI-built prototype into something you can depend on.
We also help with
Something else in mind?
If it involves data, AI or software, ask us. We'll tell you honestly whether we're the right fit.
What you get · Software that keeps working.
- Correct
- Your data stays right, even when two people edit at once or an input is wrong.
- Safe
- Clear rules on who sees what, and your data stays where you decide.
- Tested
- Changes are checked before they go live, so today's fix doesn't break yesterday's feature.
- Watched
- We hear about failures before your customers do, and there's a way back.
- Owned
- Someone is named as responsible after launch: your team with our handover, or us under an agreed plan.
Why a working demo isn't a finished product
Tools we often use: LangGraph, OpenAI API, Local LLMs, Python, Databricks, Airflow, FastAPI, Next.js, PostgreSQL, AWS
What AI can do
Where AI pays off today.
Not sure where AI fits? Tell us how you work today. We'll suggest where to start, and tell you honestly when AI isn't the answer.
How it works
From first message to delivery.
- 1
Ask the assistant
Describe what you're working on, right here. No sign-up for the first messages. Prefer email? That works too.
- 2
Sketch & estimate
Get a first solution sketch and an indicative price range in minutes. The price comes from a calculator, not the AI.
- 3
Free project plan
Want more detail? Answer a few questions and get a written plan you can edit and refine.
- 4
Intro call
A free 30-minute video call with us to talk it through and check the fit.
- 5
Proposal & build
A clear proposal in your own project room (bigger projects start with a short Discovery sprint). Then we move fast, in short iterations.
Want to see each step first? The walkthrough shows the chat, the project plan and the project room, with screenshots.
See how it worksSelected work
Recent projects.
Client names stay confidential.
- AI agents
- LangGraph
- Local LLM
This website
A chat-first site where an AI agent answers questions, sketches solutions, drafts project plans and books calls, on a self-hosted LLM with LangGraph.
Read the case- Multimodal
- AI agents
Item details and prices from photos
Multimodal AI agents that read product photos, fill in item details and estimate prices for an online marketplace, with an evaluation framework that measures every change.
- LLM
- Integrations
AI platform with custom tools
LLM chat that drafts quotes from brand templates and an internal product database, and automates marketing newsletters.
- LLM
- NLP
Customer review insights
An LLM service that summarises product reviews and scores sentiment for a global retailer, with topic models that surface what frustrates customers.
- Forecasting
- ML
Predicting media consumption
Forecasting how audiences watch across TV, streaming and online video, for audience measurement products.
Read the case- Embeddings
- Real-time
Recommendation engine
Real-time recommendations for a high-traffic online marketplace, built on image and text embeddings and proven with A/B tests.
- API
- Statistics
Real-time reference rates
Live market data from several sources, combined into one reliable reference rate behind an API.
More projects (5)
- Simulation
- In progress
Digital twin platform
A shared platform that collects data from physical systems and simulates them, to test scenarios before changing the real thing.
- ML
- Automation
Churn prediction & retention
Spotting customers who are about to leave and reaching them with personalised journeys.
- Databricks
- Azure
- ETL
Data platform pipelines
ETL pipelines on Databricks and Azure, consolidating sources into one reliable platform.
- Forecasting
- Retail
Retail sales forecasting
Sales forecasts that support stock decisions for a retail chain.
- Web app
- FastAPI
CRM application
A custom CRM web app shaped around how its users actually work, in daily use today.
Try it: ask a dataset anythingAsk a question about sample sales data. An AI agent writes and runs Python in your browser and explains the answer, with charts.
Open the data lab
Who you'll work with
Small team, senior people.
You talk directly to the people who build your solution.
Kristof Menyhert
Technical lead
Ideas, architecture, data platforms, ML and LLM engineering. Turns hard problems into systems that keep running.
Juhi Chandra
ML & AI lead · storyteller
Our storyteller in client conversations and presentations, and builds ML and AI that runs in production.
Questions
Good to know.
How do you price projects?
Every project is different, so we look at yours before we name a price. After a free intro call (and, for larger work, a short Discovery sprint) you get a clear proposal: hourly or a fixed price. We prefer project-based agreements and stay flexible for smaller projects and new collaborations. The price covers the build; hosting, usage costs and maintenance are decided and priced separately.
Do you work on a project basis?
Yes. We prefer projects with a clear outcome where we help design the solution. We don't do pure staff augmentation.
Can our data stay private?
Yes. Hosted models are used through business APIs that don't train on your data by default, or we run open models on your own infrastructure so nothing leaves your environment. The assistant on this site runs on our own hardware.
What if the AI gets things wrong?
We plan for it. Answers come from your own sources and show where they come from, we test against real questions before launch and keep measuring after it, and a person approves anything that matters, like a quote or an email to a customer.
Who owns the code?
You do. Unless we agree otherwise, the code and deliverables we build for you are yours once paid.
Will you sign an NDA?
Of course. Just send it over before we go into details.
Do you work remotely?
Online by default, which keeps things fast and affordable. We're happy to come on site for a kickoff or workshop.
What happens after delivery?
You get documentation and a handover so your team can own it. Running it afterwards (hosting, updates, fixes) is agreed and priced separately: your team can do it, or we can.
Who pays for hosting and who maintains it?
That is decided per project and priced in, not assumed. By default the infrastructure runs on your own accounts and you pay the provider directly. Your team can run and maintain it with our handover, or we can do it for a monthly fee, with hosting and usage costs on top. A mix works too, and for a local model we agree who owns the hardware.
Can you take on bigger projects?
Yes. When a project needs more hands, we bring in trusted specialists or a whole team. We stay the project lead and remain responsible for the result, so you keep one point of contact and what we promised still holds.
Tell us what you're building.
Ask a question, get a feel for scope and price range, or book a free 30-minute intro call.