Inspired by my previous work at Coupang, I attempt to rearchitect the pricing engine by building an E2E MVP for competitive pricing. The repository implements the deterministic M1 baseline, M2 adapter boundary, M3 identity mapping, M3.2 structured retrieval, and M4 operator/review workflow.
M4 operator prototype.
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
pytest -q
streamlit run ui/streamlit_app.pyOpen http://localhost:8501.
Model/shadow mapping is advisory. Only exact trusted identifiers are auto-eligible in the prototype. Production publication is not connected; publication in M4 is a mock operation.
Competitor site observation -> extraction -> normalization -> identity mapping -> market price -> pricing policy -> review -> mock publish -> audit
See docs/ for the product/engineering manual, mapping concepts, and milestone reports.
Opinions and work are my own. They do not represent my previous employer.