Research system for medical AI
Medical AI hypotheses, tested against clinical reality.
Pinah is for engineers who build medical AI without a clinical background. The Pinah Engine generates and tests research hypotheses. Shlomo keeps what the project has learned, so every round starts further ahead than the last.
Pinah Engine
From one idea to tested candidates.
The engine takes your starting hypothesis and grows it into many candidate directions. Clinical checks remove the flawed ones, and real experiments confirm the rest.
- Your starting idea
- Promising direction
- Removed by clinical check
- Confirmed by experiment
Diverge
Explore widely
From your starting idea, the engine grows a tree of candidate directions, including ones you would not have tried.
Filter
Check against the clinic
Built-in clinical rules remove ideas with hidden flaws before anyone spends time training them.
Verify
Run the real test
Surviving branches are run as experiments, and the results decide where the tree grows next.
You stay in charge. The engine pauses at key points for your decision, so the direction of the research is always yours.
Shlomo
The project’s memory.
Shlomo is where Pinah keeps what a research project has learned. It has three parts, and each has one job.
MyOntology · Defines
Shared standards
Established clinical and methods vocabularies, reused as they are instead of reinvented. They give Pinah exact definitions and the rules behind its clinical checks.
MyWiki · Remembers
Your judgment
Your research notes and decisions, turned into structured memory once you approve them. Pinah stops repeating what you already know and learns what you tend to rule out.
MyRef · Connects
The literature
The papers the project builds on, organised by dataset, endpoint, results and limits, and by how their ideas relate to each other.
Stored
Research decisions, aggregate results, paper metadata and the links between them.
Never stored
Patient records, images, or any log that identifies a person.
How they work together
A loop that sharpens each round.
The engine does the searching. Shlomo supplies what it needs to search well, and receives what the search produces.
Pinah Engine
Prepare · Explore · Verify
Shlomo
MyOntology · MyWiki · MyRef
-
Before the search
The engine reads
It pulls the definitions behind your idea, what you already know, and the closest papers, then points out what your hypothesis is missing.
-
During the search
Shlomo steers
Your earlier decisions help decide which branches grow first, and the clinical rules remove flawed ones.
-
At the checkpoints
You decide
You approve, reject or redirect. Everything the engine proposes to remember waits for your yes.
-
After the session
Memory grows
Approved entries join Shlomo. The next session starts from a stronger idea and needs a shorter briefing.
Built to change
Less guessing, more signal.
- Fewer dead endsClinical checks run before experiments, so flawed designs are dropped early instead of after days of training.
- Stronger starting pointsShlomo remembers what you have learned, so each session begins from a better question.
- Your time backThe search runs without you. You review a short list at set checkpoints.
Pinah is a research tool for engineers and researchers. It does not provide diagnoses or treatment advice, and it is designed to run on your own hardware so patient data stays with you.
Status
Early stage, with a plan to measure it.
- DoneSystem designArchitecture, data boundaries and the evaluation plan are written down and reviewed.
- In progressWorking prototypeThe Pinah Engine and Shlomo are being built on local hardware with open models.
- NextFirst validationLung cancer radiation therapy is the first domain, using public imaging data and papers published after the models’ training cutoff.
- NextClinical reviewClinical collaborators review the built-in rules before they are trusted. Every component is also tested by removing it and measuring what gets worse.
The name
The stone everything else is set against.
Pinah · cornerstone
Pinah is Hebrew for corner. In even pinah, the cornerstone, it names the first stone laid in a building. Every other stone is aligned to it.
Good research needs the same thing: a steady base under the desk and the chair, so the work on top stays square. That is the job Pinah takes on.
Contact
Talk to us.
Address copied.