Crypto data are a nightmare for FinTech ML engineers
Common wisdom states that an ML engineer spends about 80% of the time on data preprocessing and feature engineering. I buy the feature engineering part because good features are essential for a good model. However, I’m not convinced about data prepossessing anymore because I have seen the same opening ritual in nearly every new FinTech ML project: Find a quality data source. Acquire or download data. Ingest data into a DB or convert to flat files.