About
mnemic labs is an independent AI research lab in Zurich, founded in 2026. We pair research on language-model reasoning with the development of systems deployed inside partner organizations.
What we do
The research side works on faithfulness and verification: whether a model's stated reasoning matches what it actually did, and how to check each step instead of trusting the final answer.
The applied side turns that research into systems. The first is Novalis, which answers questions over an organization's documents with citations to the source passages.
How we work
We publish what we find, including the failures. We take few engagements, build against your data, and deploy inside your infrastructure. The people who did the research show up for the deployment.
What we decline
- Anything that requires your data to leave your environment.
- Work we cannot measure before handover.
- Renting out engineers by the month.
- Projects we cannot explain in one sentence.
Team
Nine people. Three lead the work.
Lena Okafor
Lena leads research. She spent six years evaluating reasoning models at a large industrial lab, and writes most of our negative-results reports.
Jonas Weiss
Jonas runs engineering and leads deployments. He spent a decade building search and indexing systems, and is usually the one inside your environment getting the first index to build.
Mara Voss
Mara runs the lab: hiring, finance, contracts, and the logistics of working inside other organizations. She came from research operations at a university.
Facts
- 2026, Zurich
- Nine people
- Research grants and applied engagements
- Papers, tools, and deployed systems
- From mneme, Greek for memory