Wildlife Graduate Assistantship Dashboard

Live dashboard: https://chrischizinski.github.io/wildlife-grad-dashboard/

Project Summary

This project tracks U.S. wildlife and natural resources graduate assistantship postings and turns raw listings into a public-facing trends dashboard.

The system is designed to answer practical questions for students and advisors:

• How many assistantships are being posted over time?
• Which disciplines are most active?
• What salary ranges are being reported?
• Where are opportunities concentrated geographically?
• How complete/clean is the underlying data?

What The Project Includes

• A scraper (src/wildlife_grad/scraper/) for TAMU wildlife job board listings.
• Analysis/classification modules (src/wildlife_grad/analysis/) for graduate relevance and discipline tagging.
• Data pipeline scripts (scripts/) to refresh and transform outputs.
• A static web dashboard (web/wildlife_dashboard.html) with JS/CSS assets and JSON-driven charts.
• GitHub Actions workflows for scheduled data refresh and GitHub Pages deploy.

Current Status (as of February 17, 2026)

This status is based on the latest local analytics artifacts in this repository.

• Analytics last generated: 2026-02-15 05:19:25 (web/data/dashboard_analytics.json)
• Dataset represented in analytics: 229 positions
• Postings with salary parsed: 138
• Current graduate-assistantship file size: 22 records (data/processed/verified_graduate_assistantships.json)
• Automation workflows present:
  • .github/workflows/scrape-and-update-dashboard.yml
  • .github/workflows/deploy-pages.yml

Project Maturity

• Core scraping and dashboard pipeline exists and is operational.
• Data products are versioned into data/, dashboard/data/, and web/data/.
• CI/CD automation is present for refresh + deploy.
• Dashboard redesign/rebuild work is currently in progress (see AGENTS.md dashboard contract).

Active Focus Areas

• Improve dashboard coherence across overview, discipline, compensation, geography, and data quality sections.
• Harden empty-state handling for all KPI cards and charts.
• Keep GitHub Pages path behavior robust (relative data fetches).
• Continue improving salary/location parse coverage and transparency.

Repository Map

src/wildlife_grad/        Python package (scraper + analysis)
scripts/                  Data pipeline and maintenance scripts
web/                      Dashboard frontend (HTML/CSS/JS + web/data)
data/                     Raw, processed, archived, and fallback artifacts
dashboard/data/           Analytics outputs used by dashboard build flow
docs/                     Supplemental architecture/setup notes

Running The Dashboard Locally

cd web
python -m http.server 8080
# then open http://localhost:8080/wildlife_dashboard.html

Optional Discord Notifications

The weekly scrape workflow can post a Discord summary after each successful run.

Configure a GitHub Actions secret named DISCORD_WEBHOOK_URL and the workflow will send:

• total graduate positions in the refreshed dashboard dataset
• salary-parsed count
• latest scrape timestamp
• titles, organizations, and locations from the latest capture batch

Continual Discipline Model Refinement

The discipline classifier now supports a versioned promoted model plus an uncertainty review queue. Detailed step-by-step guide: docs/DISCIPLINE_REVIEW_LOOP.md.

# 1) Rebuild analytics + refresh confidence queue
python scripts/generate_dashboard_analytics.py

# 2) Retrain candidate model and promote only if validation improves
python scripts/retrain_discipline_model.py

# 3) Seed conservative starter gold labels from stored assistantships, then retrain
python scripts/retrain_discipline_model.py --auto-seed-from-positions

# 4) After editing discipline_confidence_queue.csv with review decisions,
#    import them into the gold label store
python scripts/import_discipline_queue_reviews.py

Artifacts:

• Gold labels: data/processed/discipline_labels_gold.json
• Model manifest: data/models/discipline/manifest.json
• Training report: data/models/discipline/latest_training_report.json
• Confidence queue: data/processed/discipline_confidence_queue.json

Review CSV conventions (data/processed/discipline_confidence_queue.csv):

• Add review_status with one of: accept_model, keep_final, override, skip
• For override, also set reviewed_discipline
• Optional: review_notes, reviewer

Notes

• This README is intentionally project/status oriented.
• Operational command details are maintained in repo docs and script help output.