Foundation-model lab · Vienna, Austria

We train foundation models on the physics, language, and imagery of your industry.

Then we hand over the whole build — the checkpoint, the data it was trained on, and the eval report — compressed to run self-hosted on a single datacenter GPU. No third-party APIs, no per-token bill, no weights you don't own.

Request a custom build See the open checkpoints

What ships to you

Checkpoint~20 GB, quantised — runs on one L40. Yours to keep.
Training setThe corpus, geometry, or labelled frames it learned from.
Eval reportBenchmarked against frontier models on public tasks and unseen regimes.
RunbookDeployment on your infrastructure — Kubernetes, air-gapped if needed.

100% self-hosted · no third-party APIs

30K+

Hugging Face downloads

Open checkpoints published

130+

Validated simulation cases

Platforms in production

Third-party API calls

One method, three domains

Pick the line that sounds like your job.

The same method — adapt a proven base, never train from scratch — applied to medical text, to fluid dynamics, and to a factory camera feed.

Language

Private domain LLMs

Reasoning models fine-tuned on your corpus, plus the chatbots and semantic-analysis products that run on them.

84% MedQA · ~20 GB · one L40

Explore Language

Physics

Flow surrogates

Operator models for CFD regimes a foundation model never saw — without erasing what it already knows.

1.8% wake error · solver-grade data included

Explore Physics

Vision

Detectors without labels

YOLO detectors adapted to your objects, cameras, and process — trained in simulation, shipped with the labelled set.

weld 0.97 · defect 0.91 · zero manual labels

Explore Vision

Evaluated in the open

Every claim on this page is a row in a table.

Published results · Empirisch Tech checkpoints
Build	Task	Result	Adapted from	Footprint
Thinking-LQ-1.0	MedQA — medical question answering	84.0%	DeepSeek-R1-Distill-Qwen-32B	~20 GB · 1× L40
GPT-4o	MedQA — reference point	≈88%	—	Hosted API
Flow-1.0	Wake error, unseen flow regime	1.8%	Poseidon-B	Single GPU
SynYOLO — weld	Weld detection, industrial camera	0.97	YOLOX	TensorRT · real-time
SynYOLO — defect	Surface defect detection	0.91	YOLOX	TensorRT · real-time
Simvera field	Mean localisation, live factory cameras	6.8 px	SynYOLO	100% synthetic training data

◆ Our builds. The GPT-4o row is the published comparison point — Thinking-LQ-1.0 lands within four points of it, self-hosted. Weights for our rows are on Hugging Face; reproduce them yourself.

How it works

Adapt a proven base — never train from scratch.

Four steps behind every checkpoint we've shipped, whether the domain is medical text, fluid dynamics, or a factory camera feed.

Step 01

Start from a proven base

A frontier open checkpoint brings the general capability, so your data doesn't have to.

e.g. DeepSeek-R1-Distill-Qwen-32B, Poseidon, a YOLO backbone

Step 02

Adapt it to your domain

QLoRA and GPTQ for language, operator learning for physics, synthetic frames for vision — without erasing what the base knows.

your data stays on your infrastructure

Step 03

Evaluate in the open

Every build is benchmarked against frontier models on public tasks and on regimes it has never seen.

accuracy, latency, and cost per query on your hardware

Step 04

Ship the whole build

Checkpoint, training data, and eval report — compressed to run self-hosted on a single datacenter GPU.

~20 GB · no third-party APIs

Evidence, not brochures

Six platforms in production, all running on our own checkpoints.

We don't ask you to take the method on faith. These serve real users today.

Chaperone AI

chaperoneai.net

The model line itself — language, physics, and vision checkpoints, shipped with their data.

Numerical AI

numericalai.net

Three GPU solvers — CMF fluids, CEM electromagnetics, SRS turbulence. €0 licence fees.

Simvera

simvera.net

Industrial perception trained entirely in simulation, live on factory cameras.

DeepResearch

deepresearch.chaperoneai.net

Deep-research report generator built on our reasoning LLMs.

PropensityAI

propensityai.com

Search engine running on our specialised search and reasoning models.

AI 3D Generator

ai3dgenerator.fetchcfd.com

Text- and image-to-3D on a GPU-accelerated pipeline.

Find our open-source models on Hugging Face

Also on Azure Marketplace — deploy our GPU inference VMs and desktop environments directly from Azure. View offerings →

Backed by the programmes we build on

Research partners & past engagements

The practice underneath

Somebody has to run the clusters the models land on.

The delivery engineering behind every build we ship — available on its own if that is what you need.

DevOps

Pipelines, GPU scheduling, and release engineering for teams shipping models into production.

Kubernetes

Scalable, resilient clusters — including air-gapped and on-premises installs for regulated environments.

From the blog

Technical notes.

Computer vision

Image convolution and common image-processing filters

Forecasting

Stock prediction using FBProphet

Different corpus, geometry, or label set?

Book a technical call