Joshua Thomas

Multi-client field-operations platform for a political consulting firm, replacing a sprawl of Google Sheets and Apps Script.

• Built a centralized multi-client project dashboard consolidating NGP VAN, Google Sheets, Mobilize, and an internal database behind one interface, covering turf tracking, percent-to-goal analytics, canvasser quality control, and automated client reporting.
• Implemented role-based access control across five roles using bcrypt password hashing, JWT sessions, and granular per-resource permissions, so data and changes are restricted to permitted parties.
• Enforced all 19 permissions in the API independently of the interface — hiding a link is presentation, not access control.
• Implemented an audit log capturing every state-changing request and authentication event with user, IP, status, and duration. During RBAC testing it surfaced a privilege escalation where a client-facing read-only role could enumerate staff accounts.

• Python
• Parsons
• NGP VAN
• JWT
• PostgreSQL

Tracks and compares campaign finance data, with built-in compliance monitoring for contribution limits.

• Built a full-stack application that tracks and compares campaign finance data between active candidates and historical filings, on a FastAPI backend with a Next.js frontend.
• Added an internal compliance tool that continuously monitors donations and flags individual contributors approaching or exceeding legal contribution limits.
• Deployed on Cloudflare for high availability and fast delivery.

• Next.js
• FastAPI
• Python
• Cloudflare

End-to-end streaming pipeline over the Wikimedia firehose, from Kafka ingest to a live Grafana dashboard.

• Architected an end-to-end distributed streaming pipeline ingesting the continuous Wikimedia EventStreams firehose through a custom Python producer and Apache Kafka.
• Developed a PySpark Structured Streaming application to consume, parse, and transform large JSON payloads in real time.
• Engineered a custom sink with PySpark's foreachBatch to load processed micro-batches into PostgreSQL as the hot-storage layer for real-time analytics.
• Deployed a live Grafana dashboard over that database, visualizing time-series and system metrics.

• Python
• Apache Kafka
• PySpark
• PostgreSQL
• Grafana

Distributed Raspberry Pi probes measuring real-world network health across sites, feeding Prometheus and Grafana.

• Deployed Raspberry Pi nodes that expose custom Prometheus gauges for per-site request latency and node liveness, scraped every 15 seconds and visualized in Grafana.
• Wrote a threaded Python collector that probes 18 target sites concurrently with a liveness heartbeat over HTTP/2, labeling every metric by node location so a single-site slowdown is distinguishable from a network-wide outage.
• Automated node provisioning with a Bash script that installs node_exporter as a systemd service under a dedicated user and enables it at boot, reducing a new monitoring site to one command.
• Secured all inter-node data transmission using Tailscale.

• Python
• Bash
• Prometheus
• Grafana
• Tailscale

Animates six sorting algorithms side by side, with live comparison and array-access counters.

• Built an interactive visualizer for six sorting algorithms — quick, merge, and heap sort against bubble, selection, and insertion sort — with adjustable array size and animation speed.
• Instrumented each run with live comparison, array-access, and elapsed-time counters, and seeded arrays as random, nearly sorted, reversed, or few-unique to show how input shape changes real behaviour.

• TypeScript
• React