Health · GitHub · 🌐 International · via agentusecases.ai

Claude agent combines two years of WHOOP data with blood test results

An open project replaced Databricks' Genie agent with a Claude agent and measured both on 30 questions about personal health data.

Profession: Developer, Data Analyst
Tools: Claude, Databricks Genie, Databricks Agent Bricks
Source: github.com ↗
Country: 🌐 International
Found via: agentusecases.ai (Alex Finn) ↗
Report date: no date
Verified on: October 27, 2026

What they did

Two years of exported WHOOP data became tables in Databricks, and Agent Bricks extracted lab results from PDFs stored in email. The Genie agent was then swapped for a small Claude agent that writes SQL, runs the query through a read-only tool, reads the rows and queries again when needed.

How it went

The evaluation exposed two date bugs in the main table, and every agent got the affected questions wrong. After the data was fixed, Claude Sonnet went from 47% to 90% correct and Genie from 47% to 67%.

Good to know

The sample is small (30 questions) and the answers were graded by another Claude model. Lab results and wearable data are health data: give the agent read-only access and do not treat its answers as medical advice.

The prompt

I'm uploading the export from my [watch or ring] and the PDFs of my blood tests from [period]. Build a table of lab values by date, line them up with [sleep, HRV, recovery] and answer: [question]. Show the numbers behind each answer, flag anything uncertain and list 3 questions to bring to my doctor.

Replace what is in [brackets]. Prompt written by us from the account, not a literal copy of the source. Do not paste other people’s personal data (privacy laws such as LGPD and GDPR apply).

Summary in our own words. Results as stated in the original account; we have not tested or audited them. Read the source ↗ Case found via agentusecases.ai (Alex Finn); text and prompt rewritten by us. Found a mistake?