Partnerships
Partner with SoReliable.
Know a team that needs expert judgment or real-world data? Run a business where real work happens? Building robots or AI from the ground up? Researching AI or superintelligence, or a field AI is changing? We want to work with you.
- AI labs and model builders
- Robotics and embodied AI teams
- Physical AI startups
- AI agent builders
- AI researchers and universities
- Research labs and independent researchers
- Data and annotation companies
- Businesses, workshops and factories
Ways to partner
Pick the one that fits, or tell us about something else.
Introduce us
For people who know AI teams
Connect us with AI labs, robotics and physical AI companies, researchers or agent builders who need trained experts, evaluation or data. If an introduction becomes a project, we agree a referral arrangement with you in writing.
Host real-world data collection
For businesses, workshops and factories
Robots learn from people doing real work. With your permission, our trained members record first-person video of everyday tasks at your site. They follow your safety rules and keep customers, faces and private documents out of view. You agree the terms before anyone films.
Build with us from day one
For robotics and AI companies
Starting a new robot or AI model? We can run the whole data side for you: decide what it needs to learn, find and train the right people, capture the data, label it, check it and deliver it ready for training.
Research with us
For researchers, universities and research labs
Join our research on superintelligence and reliable AI, or bring us into your own research in any field. See the research section below for the ways we can work together.
Research with us
We study superintelligence and reliable AI because the results shape what we build. Research is better done together, so there are several ways in.
Research track
Superintelligence
What it means, how close it is, and how people can keep judging work that may surpass their own.
- Fd10Foundations
- Cp11Capabilities
- Al12Alignment
- Ov13Oversight
- In14Interpretability
- Gv15Governance
Research track
AI research
How expert judgment makes AI reliable: evaluation, agents, deep research, rubrics, and AI in clinical, software and robotics work.
- EvEvaluation
- AgAI agents
- DrDeep research
- RuRubrics
- ClClinical
- RbRobotics
- i
Join our superintelligence research
Work with us on the open questions in our superintelligence track: measuring expert oversight of work harder than the overseer could do, tracking how fast AI is speeding up AI research, and what an independent audit of a frontier system should test.
Start a conversation - ii
Join our AI research
Build benchmarks, rubrics and studies with us on how human experts and AI systems judge the same work, in software, clinical, legal, science and robotics tasks.
Start a conversation - iii
Include us in your research
Researching medicine, law, education, agriculture, language, climate or any other field AI is changing? We can design the human side of your study, supply trained expert reviewers and human baselines, collect real-world data, and measure agreement between reviewers.
Start a conversation - iv
Write and review with us
Co-write an article for our research tracks, or review a draft in your specialty before it is published. Every article lists its sources.
Start a conversation
Trained experts on call
Reviewers in many fields, assessed and trained before they touch your study.
Measurement built in
Rubrics, gold answers and agreement between reviewers, measured rather than assumed.
Credit agreed in writing
Authorship, acknowledgement, data ownership and publication rights are settled before the work starts.
Ethics and consent
Participants agree to how their work is used, and we follow your ethics approval where one is needed.
From an idea to training-ready data
For robotics and physical AI teams, we can handle every step, or just the ones you need.
- 1Sc
Scope
We agree the tasks, settings and quality bar with you, in writing, before any work starts.
- 2Ex
Experts
We find people with the right hands-on skills and train them on your standard.
- 3Ca
Capture
Members record first-person video of real work on phones our app has checked for 1080p at 60 fps.
- 4La
Label
Trained annotators mark each step, action and object, frame by frame.
- 5Qa
Check
Every clip is reviewed against your checklist. Clips that miss it are redone, not delivered.
- 6De
Deliver
You receive clean, documented data in the format your training pipeline expects.
What every partnership includes
A written agreement first
Scope, data handling, ownership and payment are agreed before any work begins.
Consent and privacy built in
No faces, customers or private documents on video unless everyone involved has agreed.
Quality you can check
Clear acceptance rules, and a report on what passed and what was redone.
A direct line
You work with the people running your project, not a ticket queue.
Let's talk about working together
Tell us who you are and how you'd like to partner. We read every request.