Clinical ML reporting checklist that survives reviewer pressure
A practical checklist for clinical prediction and imaging ML so results are interpretable, calibrated, and defensible.
Independent Research Infrastructure
Ideas to funded proposals, reproducible evidence, and publication ready outputs
Theals is an independent research infrastructure practice supporting grants, large scale data work, applied machine learning, and publication ready writing. We deliver defensible methods, clear assumptions, and submission ready artifacts across the full research pipeline.
Led by a doctoral trained team spanning clinical medicine, data science, and scientific publishing.
Divisions
A single practice organized into clear divisions, designed to move work from question to publication.
Grants and strategy
Shape fundable questions, align aims to reviewer logic, and build a proposal that is coherent end to end.
Data and evidence
Cohort definitions, pipelines, QA, and analysis plans that make results reproducible and auditable.
Statistical analysis
Inference with explicit assumptions, sensitivity checks, and outputs that withstand scrutiny.
Machine learning
Evaluation that matches the decision, with attention to labels, bias, drift, and interpretability.
Research writing
Manuscripts, methods, and results built to survive peer review and reduce revision cycles.
Publishing and packaging
Submission ready tables, figures, responses to reviewers, and dissemination focused outputs.
Use cases
Representative engagements that show how the practice operates across grants, data, models, and publication.
Grants and strategy:
From idea to fundable proposal
Example:
Grant and study design build. Clarify the question, endpoints, and feasibility. Produce a proposal narrative, analysis plan, and a realistic execution path.
Typical deliverables:
Specific aims and narrative structure
Study design and analysis plan
Power and feasibility notes
Revision mapping to review criteria
Data and evidence:
From raw files to defensible cohorts
Example:
Large scale clinical data build. Convert raw extracts into a cohort, standardized features, and analysis ready tables with versioned logic and QA.
Typical deliverables:
Cohort definition and phenotype logic
Reproducible ETL and data dictionary
QA checks and audit trail
Analysis ready tables and summaries
Machine learning:
From prototypes to credible evaluation
Example:
Model development and evaluation. Build or validate ML models with decision aligned metrics and clear failure analysis.
Typical deliverables:
Labeling strategy and baseline models
Validation plan and reporting
Error analysis and interpretability notes
Model card and deployment considerations
Research writing and publishing:
From results to submission ready outputs
Example:
Manuscript and publication packaging. Convert analyses into a coherent paper, figures, and reviewer ready revisions.
Typical deliverables:
Manuscript drafting or restructure
Methods and results synthesis
Figure and table logic
Response to reviewers and resubmission plan
Frequently Asked Questions
What kinds of clients do you work with?
Academic labs, clinical research groups, biotech and startups, and publisher adjacent teams needing rigorous methods and clear scientific outputs.
Do you take on end to end projects or only one part?
Both. Engagements can be scoped to a single division, or run end to end from grant through publication packaging.
How do you keep work reproducible and auditable?
Clear cohort definitions, explicit assumptions, versioned outputs, and QA checks designed to make results defensible and easy to revisit.
Do you provide writing without data work?
Yes. Writing engagements focus on structure, claims discipline, and reviewer facing clarity, including revision strategy and responses.
Contact
Send a short brief. You will receive a scoped response with next steps.
Email
Scope
Grants, data, ML, writing, publishing
Response
Scope, timeline, and terms
Insights
View all posts »Notes on research infrastructure: grants, data, methods, modeling, and publication.
Project intake tracks and what to send first
Pick a track, send the minimal inputs, and get a first pass output quickly without scope drift.
Writing revisions that increase clarity without changing meaning
A repeatable editing pattern for scientific writing that reduces reviewer confusion and strengthens causal logic.