An end-to-end data product that ingests U.S. federal contracting data, resolves messy vendor entities, and surfaces explainable scores for recompete likelihood, incumbent strength, and market concentration across agencies and NAICS codes.

MVP slice: Department of Homeland Security × NAICS 541511/541512 (IT services) × FY2020–2025.

Surfaces intelligence that commercial tools (GovWin, Bloomberg Government) charge $20K+/seat/year for — from free public data.

.
├── apps/web/              # Next.js 14 app router UI
├── api/                   # FastAPI service
├── pipelines/             # Python ingestion + load scripts
├── dbt/                   # dbt project (raw → staging → marts)
├── data/
│   ├── raw/               # Local parquet cache (gitignored)
│   ├── reference/         # NAICS/PSC/agency seed data
│   └── vendor_manual_merges.csv
├── mocks/                 # Static HTML design references
└── .github/workflows/     # Ingest + CI

cp .env.example .env
# Fill in DATABASE_URL once you have a Neon instance

cd pipelines
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

# Pull ~5 years of DHS IT-services awards (~100–300k rows, ~5–15 min)
python ingest_usaspending.py \
  --agency-code 070 \
  --naics 541511 541512 \
  --fy-start 2020 --fy-end 2025

# Load local parquet into Postgres
python load_to_postgres.py

cd dbt
cp profiles.yml.example ~/.dbt/profiles.yml  # edit as needed
dbt deps
dbt run
dbt test

cd api
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
uvicorn main:app --reload
# http://localhost:8000/docs

cd apps/web
npm install
npm run dev
# http://localhost:3000

(Agency, NAICS, PSC, Time) is the market cell. Vendors hold share in each cell. Awards are the evidence. All analytical views project from this tuple.

Core entities: agency, vendor (resolved), vendor_alias, award, award_transaction, naics_code, psc_code.

Every score is transparent SQL with a per-component breakdown surfaced in the UI. No black-box ML.

• Phase 0 — Foundation (initial scaffold)
• Phase 1 — MVP with scored radar + vendor + agency views
• Phase 2 — SAM.gov opportunities, multi-agency, CI, demo video
• Phase 3 — Forecast scraping, vehicles, exports

MIT — data derived from public U.S. government sources (USASpending.gov, SAM.gov).