Hasna Foundation
AI research for humanity's hardest problems.
We exist to fund open research on health, climate, education, and access to science, and to judge it by what it makes possible for people.
Why the foundation
Most AI research follows the product cycle: what can ship next quarter, what a benchmark rewards, what a customer will pay for. Some of the most important questions do not fit that rhythm. How does a clinic with one doctor and unreliable power use a model safely? How does a coastal city plan for a climate it has not seen? How does a student with a shared phone keep learning?
Hasna Foundation exists to make room for that work. Our role is to support researchers close to the problem, publish results so others can check and reuse them, and stay involved long enough to learn whether an approach actually helped.
We are not a product company. Our measure of success is what the research makes possible for people who will never meet us.
Focus areas
Where the work goes
Six areas where careful AI research can change what is possible for people. Each one is judged by what it changes outside the lab.
Reliable AI for science
Methods that help researchers trust what a model tells them: calibrated uncertainty, provenance, and evaluation that holds up outside a benchmark.
Care where it is scarce
Decision support for clinics with too few staff, intermittent power, and limited connectivity. The constraints are the design brief, not an afterthought.
Climate and resilience
Models and data tools that help communities anticipate heat, flooding, and crop failure, and decide what to do before the season arrives.
Access to learning
Tutoring and teaching tools built for crowded classrooms, shared devices, and the languages students actually speak.
Open tools for discovery
Shared datasets, evaluation suites, and software that lower the cost of careful research for groups without large compute budgets.
Technology that includes people
Interfaces and models built with disabled users from the start, rather than adapted for them after launch.
Method
How we work
01
Research grants
Direct funding for small teams working on a well-defined problem, sized so the work can be done carefully rather than rushed.
02
Open research
Results, code, and data are published so others can check, reuse, and extend them. Negative results belong in the record too.
03
Fellowships
Time and support for researchers who want to pursue a public-interest problem without leaving their field or their community.
04
Partnerships
Long-term collaboration with universities, clinics, schools, and community organizations that understand the problem firsthand.
Research
The agenda
Four areas where careful AI research can change what is possible.
01
Reliable scientific AI
Scientific tools built on AI need to say when they do not know. We are interested in uncertainty, provenance, and evaluation that scientists can actually use.
02
Climate modeling
Faster, finer-grained models and better ways to turn climate data into local decisions about water, heat, and land use.
03
Low-resource healthcare
Decision support that assumes scarce staff, unreliable power, and limited connectivity, rather than the conditions of a well-resourced hospital.
04
Education access
Teaching tools that work in large classes, on shared devices, and in the languages students actually speak.
Programs
Programs
What we are building, and where each program stands.
- Planned
Open Research Grants
Small grants for teams working on one of the four agenda areas. The first call has not opened yet; details will be published here when it does.
- Planned
Foundation Fellowship
Support for researchers who want to work on a public-interest problem while staying connected to their field. Fellowship terms are being developed.
- Planned
Research Access Fund
Support for travel, data, compute, and publication costs, the small expenses that too often decide whether a project continues.
How we decide
Principles
- 01
Public benefit is the test
We ask what a project makes possible for people, not how impressive it sounds in a summary.
- 02
Open by default
Methods, code, and results are published unless there is a specific, stated reason not to.
- 03
Rigor without gatekeeping
Careful work can come from anywhere. We review ideas on their merits and publish the criteria we use.
- 04
Humility about what AI can do
We are skeptics as well as advocates. Some problems will not yield to these tools, and saying so is part of the work.
Research notes
Occasional updates on grants, results, and open questions from the foundation. We write when there is something worth your time.
Bring us a problem worth working on.
We want to hear from researchers, clinicians, educators, and organizations that live close to the problems in our agenda. Tell us what you are working on and what would make the next step possible.
Start a conversation