Models can propose
Frontier models generate hypotheses and experimental plans good enough to be worth running at scale.
About
Removing the human bottleneck from discovery
AutoScientia is an autonomous AI company. We build systems that run the full experimental cycle — propose, run, evaluate, iterate — so that scientific and engineering progress stops being limited by how many experiments a person can personally shepherd.
Mission
Automate the method, not the answer
Ideas are not scarce. Testing them is slow, and it is slow for a boring reason: the testing is sequential, manual and expensive in attention. Design, run, wait, record, decide, repeat. Each turn of that loop costs days or weeks, and working harder does not make it shorter.
Our answer is to build AI-enabled systems that close the loop themselves — using frontier models to propose and large-scale compute to run experiments in parallel — then expand that capability from machine learning research outward to problems where the same loop shape decides how quickly humanity makes progress.
At a glance
Founded2025
HeadquartersGermany
FocusAutonomous experimentation
First domainML research & engineering
Websiteautoscientia.com
A small, full-stack team working across chips, infrastructure, models and end-to-end products.
Why now
01
02
03
04
The pieces finally line up
Automating a scientific loop needs three things to be true at once, plus a fourth that holds them together. All four arrived recently.
Compute can run
Large-scale parallel infrastructure turns a thousand experiments from a decade into an afternoon.
Results can be scored
Automated evaluation makes a loop trustworthy: measurable objectives instead of opinions.
One stack can hold it
Silicon, infrastructure, models and products are close enough to build as a single system.
Founder
Who started AutoScientia
Adam Loewy
Founder
Adam Loewy founded AutoScientia in 2025 in Germany, around a single observation: the scientific method itself is the bottleneck. The tools researchers use have improved enormously, but the loop they work inside has barely changed — one experiment, one answer, one decision at a time.
The company exists to rebuild that loop as a system: autonomous experiments that propose, run, evaluate and iterate without a human standing in the middle of every turn. Adam leads AutoScientia's technical direction and its full-stack approach — from compute and infrastructure through models to the products built on top.
AutoScientia is at the start of that work. The first target is the domain where progress can be measured most honestly: automating machine learning research and engineering, using the results to improve the very stack the loop runs on.
How we work
Principles we are willing to be judged against
- Measure, then claim. If an improvement cannot be scored, it is a hypothesis about our own work, not a result.
- Automate the boring loop. Human attention belongs on objectives, not on babysitting runs.
- Own the whole stack. Progress in one layer should pay for progress in the next.
- Be our own first customer. We point the loop at our own stack before asking anyone else to trust it.
- Stay honest about limits. Autonomy is a direction. We say where we are on the way.
Company card
Since 2025NameAutoScientia
Domainautoscientia.com
CountryGermany
FocusAutonomous experimentation
Work on the loop that speeds up everything
We are a small team in Germany, building in the open where we can and hiring where it counts.