I build the analytics, reporting and AI layer at the National Fish and Wildlife Foundation — two years there, after four years in network operations at Cisco.
dbt · BigQuery · Terraform · Python · SQL · GCP · AWS
Remote · US Eastern · open to analytics / data engineering roles
Ingestion through to a modeled, served layer — provisioned as code and running on a schedule, not by hand.
A personal data lake and knowledge graph on Google Cloud, provisioned end-to-end in Terraform and deployed from GitHub Actions on every push.
3 ingestion pipelines · 11 GCP services · BigQuery raw → knowledge · daily Cloud Scheduler runs · 38 commits
Terraform · BigQuery · Cloud Run · dlt · GraphRAG · Workload Identity Federation · GitHub Actions
`terraform apply` takes it from an empty project to a running lake — IAM, secrets and schedulers included. No console clicking.
Governed, read-constrained interfaces that let a language model query a data store without being handed the keys to it.
An MCP server over a normalized SQLite database of recipes, inventory and cooking decisions. Ranks what you've already saved by how little you'd have to buy.
18 MCP tools · 8-table normalized schema · 7 test modules · CI on every push · MIT
Python · SQLite · MCP · pytest · GitHub Actions
The interesting part isn't recipes — it's unit normalization and ingredient identity, which is where this class of tool quietly fails.
A tutor that can't skip ahead. It decomposes a topic into a prerequisite graph and refuses to release the next element until you've demonstrated mastery of the ones it depends on.
7-table prerequisite graph · SM-2 spaced repetition · 5 test modules · Docker · CI · MIT
Python · SQLite · MCP · SM-2 · Docker · GitHub Actions
Sequencing stops being a polite request in a prompt and becomes a data dependency the server enforces.
Teaching the stack, and building the tool when the off-the-shelf one couldn't express the query.
A free hands-on course teaching dbt on real National Park Service biodiversity data, plus a read-only MCP server for querying the finished warehouse in plain English.
5,824 species · 23,000 observations · 4 staging + 3 mart models · 6-step interactive lab · MIT
dbt · DuckDB · Python · MCP · Killercoda
Real public data with real messiness, not a toy `jaffle_shop` clone — and the lab runs in a browser with nothing to install.
A crawler that reads companies' own applicant tracking systems rather than aggregators, then filters on what actually disqualifies a role — seniority, geography, and stated years of experience.
7 ATS platforms · 438 live boards · 27,140 postings · zero dependencies
Python 3.11 standard library only — no requests, no scraper, no framework
Most job boards are a thin view over the same handful of ATS APIs. Read those directly and every row is a live posting with a verifiable requirement.
Case study → private repo
B.S. Statistics & Analytics, Second Major in Computer Science
University of North Carolina at Chapel Hill · May 2020