Hi Devs,
I'm Preston Marwicke, Fund Manager at Velthorne Asset Management. Usually, I talk about P/E ratios, but today I want to talk about our tech stack.
In 2026, you can't trade emerging markets with just a Bloomberg terminal. You need code.
We recently deployed a custom Python scraper to analyze the sentiment around Brazil's "Pix" payment system and Open Finance adoption.
The Challenge: Official data lags by 30 days. We needed real-time signals.
The Solution: We built a Natural Language Processing (NLP) pipeline using:
Scrapy to ingest data from Brazilian financial news portals (local Portuguese sources).
LLMs (Fine-tuned) to interpret the "Tone" of regulatory announcements from the Central Bank (BCB).
Pandas/NumPy to correlate "Tone" with historic intraday volatility of Fintech tickers like $NU.
The Result: We found a 0.82 correlation between positive regulatory sentiment in local dev forums (discussing API stability) and subsequent stock performance. Basically, if the devs are happy with the Open Finance APIs, the stock goes up.
At Velthorne, we believe the future of Asset Management is Human + Code. We don't replace traders with bots; we give our traders bionic suits.
If you are a dev working in Fintech or Data Science, I’d love to hear how you handle multi-lingual sentiment analysis.
Top comments (0)