Sam Celarek Sam Celarek Healthcare Data Analyst

Former Neuro ICU Nurse · Portland, Oregon

  • 3M+ IVF cycles in a database I designed
  • SQL Production SQL Server, Python and pipelines
  • 41% fewer central line infections after a QI rollout I led
My Résumé LinkedIn GitHub Baby Steps Book a Call Email Me

Data Analyst · Oct 2023 – Oct 2026 · Portland, OR

Nerevu Group

Production data work for a small Portland firm

Pipelines that run every night, tests that keep them honest, and reports clients act on.

30 → 1minutes of lead prep
3×lead volume
18pipeline steps, sole programmer
1,000+applicants in a funnel analysis

⚙️Government contracting pipeline

I was the sole programmer on an 18-step Python, PostgreSQL and Airtable pipeline on Heroku that turns raw government contracting data into ready-to-use leads.

CollectREST APIs, web scraping, PDF and XFA extraction
→
EnrichLLM summarization, with an automated scorer grading output quality
→
StorePostgreSQL, synced to Airtable for the team
→
UseLeads ready in about a minute

⏱️Lead prep time

BeforeBefore: 30+ minutes per lead30+ min AfterAfter: about 1 minute per lead~1 min
Minutes to prepare one lead.

📈Lead volume

BeforeBefore: baseline volume1× AfterAfter: three times the volume3×
Leads handled, relative to before the pipeline.

🧪Fixes that stay fixed

I set up the project's error logging and its first CI. Every production error followed the same loop, so a problem couldn't quietly come back.

1 · Error loggedProduction error is captured automatically
→
2 · IssueIt becomes a tracked issue
→
3 · TestA test that recreates it is written first
→
4 · FixThe fix ships with that test pinned in CI

🗄️SQL Server and SSIS

Maintained nightly SQL Server and SSIS loads for a mortgage lending client:

  • Stored procedures
  • Table-comparison scripts across loan tables
  • SQL Server Agent alerts on the nightly jobs

🔍Analysis and model evaluation

  • Hiring funnel: analyzed a 1,000+ applicant funnel stage by stage to model which candidates advance.
  • Classifier evaluation: designed accuracy tests for a classifier that filled in missing demographic data, replacing a 30-minute manual review per record.
  • LLM quality: wrote an automated scorer that grades LLM output inside the pipeline.

📝Reports

Produced the charts and edited the final documents for a municipal data governance report and a market disparity study, including colorblind-safe visuals and an executive recommendations table.

🧰Tools

  • SQL Server (T-SQL)
  • SSIS
  • SQL Server Agent
  • PostgreSQL
  • Python
  • Airtable
  • Heroku
  • GitHub Actions
  • pytest

← Back to portfolio