What the AI does
Language models help structure the question, infer what the data means, recommend methods and interpret results.
Axiom
Axiom: From clinical data to research evidence.
Overview
Your question, structured.
The clinical question is broken into population, exposure, comparator, outcome, timeframe and covariates. Your original wording is kept as you wrote it, next to progress and the next action.
Data
Upload, then confirm.
Upload a CSV or Excel file, or generate a synthetic population. Axiom infers what each field means, its units, identifiers, candidate outcomes and possible PHI, and shows it to you to confirm.
Explore
See the data before you test it.
Profiling and charts, plus watchers: computed observations about data quality, anomalies, potential confounders and bias.
Analyze
Cohort, test, model.
Set cohort rules and transformations. Axiom recommends a statistical test with its rationale and assumption checks, and trains models with guards against leakage.
Findings
Evidence against your question.
Results are written up against the original question, with limitations and provenance on every card, and exported as a PDF report.
The rule that governs everything
AI assists. Computation is authoritative.
Where every number comes from
Every number comes from standard statistics libraries: scipy, statsmodels and scikit-learn. A language model never produces a p-value, a coefficient, a metric or a cohort count. Guards in the product enforce this, and they fail closed.
Change the data, and everything derived from it is marked stale rather than left silently wrong.
Early access
Bring a question and a dataset.
Tell us about your research and we will reply personally about getting you started with Axiom.