How 781 outlets frame the climate crisis

Technical Lead & ML Engineer · Feb – Jun 2026

macro F1 in French

0.82

macro F1 in French

articles classified

~20k

articles classified

place, Environmental Award

2nd

place, Environmental Award

QuotaClimat, a French media watchdog, files evidence with the regulator ARCOM. I led a team of five building the tool behind it: a scraper, a classifier pipeline and a dashboard.

At a glance

• GPT-5-mini labelled the corpus. We checked those labels against human annotators and distilled them into multilingual XLM-RoBERTa classifiers that beat the partner's keyword baseline.
• Macro F1 0.82 in French, within ~11% of that in the other languages tested. Calibrated per frame, so a score means what it says.
• Owned the architecture: a session-parallel scraper (~2k articles/hr), Postgres, FastAPI, a Next.js dashboard, Docker. 2nd place, Environmental Award.