Preston Marwicke, fund manager at Velthorne Asset Management, says trading Brazil’s emerging market now demands more than a Bloomberg terminal. His team rolled out a custom Python scraper that harvests Portuguese news about the Pix payment system and Open Finance rollout, feeding raw text into an NLP pipeline for analysis.

Official data lagged thirty days, forcing the team to chase real‑time signals. They employed Scrapy to pull articles from local portals, then applied fine‑tuned LLMs to score regulatory tone from the Central Bank (BCB). Pandas and NumPy linked sentiment scores to intraday volatility of fintech ticker $NU, revealing a 0.82 correlation.

Velthorne frames the approach as “Human + Code,” giving traders a bionic suit rather than a black‑box bot. Multi‑lingual sentiment pipelines could become standard for Latin‑American equities, where official releases are slow. Watch for broader adoption of fine‑tuned LLMs in real‑time regulatory monitoring across other emerging markets in the near future.