Data and ethics
What the work runs on, where it runs, and what leaves the room.
Sources
De-identified research cohorts under signed data use agreements, and hospital cohorts under institutional ethics approval. Each published record names its cohort. Ethics approval numbers are given on request.
Where analysis runs
On a university high-performance computing cluster. Participant-level data never leaves it. This site, its figures, and the public tools carry aggregates only.
Small cells
On this site and in current public exports, any count below five is suppressed.
Code
Public tools are released under BSD-3-Clause or MIT with a software record on Zenodo. For new work, the analysis pipeline is released with the paper where data-use agreements allow.
Clinical software
This public bio is separate from clinical software. AutoVS and the LINE intake use separate clinical workflows. Every write needs an explicit approval and a read-back check.
This site
Hand-built in Python; one module renders every page from a few data
tables, and publications.bib is the source of
truth for every record. Served from a Cloudflare Worker, HTTPS only, with
optional first-party aggregate analytics only after your permission, and no third-party scripts. Privacy and browser storage settings explain the choices. IBM Plex is self-hosted
under the SIL Open Font License. Every figure is drawn from published
aggregates. A readme for language models is at
/llms.txt.
Requests
Data questions and collaboration requests: me@dhupbh.com. If you think a figure on this site shows too much, tell me and I will take it down first and argue second.