Artifacts

pyaar harmonize

pyaar harmonize turns messy healthcare data into analytics-ready data, a growing suite of vocabulary and data-quality harmonizers aligned with the open-source Tuva Project. The first tool, drug-name normalization, maps brand names, abbreviations, and misspellings to standardized RxNorm generic names, collapsing hours of manual medication mapping into a two-minute, privacy-first (in-browser) process.

FDA Devices MCP Server

An MCP server that gives AI assistants structured access to FDA medical device databases, 510(k) clearances, PMA approvals, adverse events, recalls, and device classification. Built with TypeScript, includes synonym expansion and source links.

Unlocking Healthcare Insights: First Look at Snowflake's Synthetic Dataset

Snowflake released a Synthetic Dataset that could be a nifty way to test out POC’s for searching through structured EHR Patient Data

SaMD Clearance Visualization

Interactive visualization of FDA 510(k) clearances for Software as a Medical Device (SaMD). Explores regulatory trends, predicate device networks, and clearance timelines across product codes.

Optimal Vascular Access for Hemodialysis Patients

When’s the best time to start a patient on hemodialysis? Too soon could lead to costly revision surgeries, too late could risk infection. Nephrologists and patients need more data to understand the schedule for a patient starting hemodialysis

Clinical Data Stewardship: A Rant

“Why are these column values null?”

FDA CRL Analyzer

Analyzes 202 FDA Complete Response Letters (2020-2024) to extract regulatory intelligence, deficiency patterns, therapeutic area trends, and sentiment. Uses DuckDB for fast SQL analytics and Parquet for efficient storage. 8 deficiency categories, 7 therapeutic areas.

Make Health Tech Great Again

More than 60 companies pledged to work collaboratively to deliver results for the American people. This database is tracking the efforts and recent achievements of each of the organizations. The apps will assist in the delivery of key services to beneficiaries including: Diabetes and obesity management, the use of conversational AI assistants to help patients check symptoms, navigate care options, schedule appointments, and among other tasks; and tools to “kill the clipboard” by replacing paper intake forms with seamless digital check-in methods

Ways to Disrupt Healthcare

musings on healthcare problems

TidyTuesday Agent

An AI agent that autonomously answers data analysis questions about TidyTuesday datasets. Uses a 5-step workflow: understand the question, load data into DuckDB, run SQL queries, interpret results, and format answers. 61 tests, domain-aware question interpretation.

Innolitics Medtech Newsletter

Two years writing and growing Innolitics' biweekly Medtech newsletter on FDA regulatory strategy for AI/ML Software as a Medical Device (SaMD). Grew it to 2,500+ subscribers, with 510(k) breakdowns, predicate analysis, AI/ML guidance, and submission lessons for founders.

pyaar-dicom

A local-first DICOM annotation station for crowdsourcing radiologist validation of medical imaging datasets. Pairs Apple's newly DICOM-approved Studio Display XDR with Kitware's open-source viewer, the public sandbox runs on FDA-released synthetic mammograms; the local install keeps real PHI on the radiologist's own machine.

pytrials-v2

A modern, fully-typed Python SDK for the ClinicalTrials.gov API v2: Pydantic models, async pagination, a validating query builder, rate limiting, and DataFrame integration. Aims to be the default Python library for clinical-trial data.