About the lab
Tunedness is an applied AI research and engineering lab. We are hired when a model almost works and nobody can say why it does not — when the prototype demos well and fails in the hands of real users, when the eval score stopped moving three sprints ago, when the agent works until the day it does not.
What we actually do
Most of our work sits in six places: model tuning and alignment, retrieval and knowledge systems, agent engineering, evaluation and observability, inference and cost engineering, and longer embedded research partnerships. Which of those a project needs is a question we answer by reproducing the failure ourselves, not by asking you to pick from a menu.
The first deliverable of every engagement is a written diagnosis: what is actually wrong, how we know, and what it would take to fix. It is yours whether or not we build the fix.
How we work
We stay deliberately small, so the person answering your question is the person who ran the experiment. Three rules hold that together:
- Research in the open loop. Weekly written updates, failures included. You see the eval numbers the same day we do.
- No account layer. The people on the first call are the people writing the code.
- You own the artifacts. Weights, prompts, eval sets, infrastructure code, runbooks. We leave holding nothing that you need in order to keep going.
Research
Roughly the same discipline applies to the questions client work keeps raising. Three of them are permanent: alignment under domain shift, evaluations that survive production, and competence per parameter. Notes and preprints are published as they are written.
The particulars
- Founded: [ YEAR ]
- Where: [ CITY ], and remote
- Team: [ N ] researchers and engineers
- Contact: hello@tunedness.com
The bracketed values above are placeholders. Replace them before launch.