Generative AI & LLMs
Custom assistants, RAG systems and fine-tuned, locally-hosted language models on your own data.
Our AI Lab · Amsterdam
World-class builds need world-class compute.
Our Amsterdam lab runs a full Apple Silicon stack, the latest and most powerful chips Apple makes. That means we can train, fine-tune and serve large models entirely on-device, with the kind of unified memory that lets a single machine hold models most setups cannot.
No cloud lock-in. No data leaving the building. No per-token surprises. Just fast iteration and production systems engineered to ship.
AI Lab · Amsterdam
The stack
A full Apple Silicon arsenal.
The most powerful chips Apple builds, chosen for massive unified memory and on-device training. Among them:
M3 Ultra Flagship
Apple's most powerful silicon, with up to 512GB of unified memory for the largest local models and training runs.
M5 Max Newest
Our latest-generation Max silicon, with a next-generation Neural Engine tuned for fast large-model inference.
M4 Pro
An efficient powerhouse for training, fine-tuning and rapid prototyping across the lab.
M4 Max
High-bandwidth GPU built for vision and generative workloads.
M2 Ultra
Proven workhorse with up to 192GB unified memory for heavy pipelines.
M3 Max
Balanced performance for production inference at scale.
…and more across the lab. A full Apple Silicon fleet, kept current with Apple's newest releases.
Massive unified memory
Run and train large models that simply will not fit on conventional GPUs.
Private by design
On-device training and inference, so sensitive data never leaves the building.
Performance per watt
Silent, efficient, sustainable compute that runs heavy workloads cool.
Local, no lock-in
Train and serve locally, with no cloud dependency and no per-token bills.