jared@platform:~$ whoami

Jared Sloan

Senior Data Platform Engineer
current company
CVS Health (remote, New Haven, CT)
background
Web developer and data analyst → SQL Server DBA in insurance (24/7 production on-call) → multi-cloud data platform engineer
education
MS, Communication, Central Connecticut State University · BS, Information Technology, UMass Lowell

I design and run SQL Server and PostgreSQL platforms on Azure, AWS and GCP, built with Terraform and Ansible and monitored in Grafana. Claude Code and GitHub Copilot write much of the scaffolding and migration code; I review it and own what ships.

Stack

Databases & pipelines

  • SQL Server, Azure SQL Managed Instance
  • PostgreSQL, PostGIS
  • AWS RDS, Oracle, MySQL
  • SSIS, Azure Data Factory, ETL/ELT

Cloud & infrastructure

  • Azure, AWS, Google Cloud
  • Terraform, Ansible
  • GitHub Actions, Jenkins

Code & AI tools

  • Python, PowerShell, Bash
  • Claude Code, GitHub Copilot

Monitoring & docs

  • Grafana, Prometheus, SolarWinds, SSRS
  • Mermaid, Lucidchart

Products

Four live products I designed, built, and run.

42EVs modelled
EV fast-charge time and highway range from each car's own charge curve, speed and temperature. One physics model feeds the site, the API and 42 vehicle pages, scored in CI against 16 published road tests it never saw: median error 8.8%.
Limits: 38 of the 42 curves are estimates from published charging reports, and the worst road-test miss is 34%. Every result says which kind it is.
225,000parcels, 18 towns
Connecticut property records joined at the parcel in PostGIS: flood zones, sales back to 2001, wetlands, septic, sewer and water, in one report per address. The market estimate shows its working. Runs on one free-tier VM with nightly backups and a restore CI tests weekly.
Limits: state sales data is annual, so most towns' newest sales are months old, and some towns publish no utility maps. Each report says so.
8towns live
Town planning and zoning portals (PDF agendas, scanned notices, messy HTML) turned into one searchable map with a public API and an RSS feed per town. Each town's parser is tested against documents that town published.
Limits: eight of Connecticut's 169 towns so far; a town goes live only after its parser runs against the live portal.
Shopsmall business
The website for Mineral Saga, my mineral business: fluorescent and other Connecticut specimens. A simple static catalogue on S3 and CloudFront, like this page.

How it runs

The shared deployment, data, and security setup behind the products.

deploys
Dev first. A deploy counts only when the health endpoint reports the shipped commit and a browser check passes; prod keeps timestamped releases, so rollback is a symlink swap.
identity
No long-lived cloud keys: GitHub Actions reaches GCP, AWS and Azure through OIDC. SSH goes through Identity-Aware Proxy; port 22 is closed to the internet.
data
Reloads land in staging, pass row-count checks, then swap in one transaction. Nightly dumps go to a second cloud, and CI restores one every week to prove it works.
CI
Lint, type checks and tests against a real PostGIS container on every push; secret scanning; actions pinned to commit SHAs; uptime checks on every site and API.

Example: fresher sales data, at no cost

Connecticut's 2026 sales release skipped Branford and Wallingford, so Deedective's reports there stopped at September 2024. A weekly crawl of the towns' assessor sites now loads about 2,000 newer sales across three towns, each swapped in one transaction and refused if it shrinks by half. It first ran in GitHub Actions and used the month's free minutes on day one, so it moved to the database VM, where it costs nothing and GitHub has no path to the database.