During the previous crypto craze a few years ago I built a .net service that continuously scanned 100s of sites for crypto news for specific keywords (configuration of sites and keywords stored in xml files) and texted me alerts when keywords hit, in an attempt to apply an event-based investment strategy to the crypto space. Given the immaturity of the market, small events often led to irrational spikes, and I made some money on quickly acting before the news became too widespread.

I started getting too many alerts so I built a machine learning based system to weed out uninteresting/unrelated stories (basically a spam filter, was 98%+ effective). By this time the service had a database of 100,000+ headlines for crypto stories, so I had good data for the filter. I built a UI to quickly allow me to train the ML algorithm by manually rating stories as good/bad and give me the ability to train/retrain/enhance the models on an ongoing basis as my crypto interests changed.

When the market crashed I lost interest. Was already full time on something else as well. I had visions of deploying the system publicly as a SaaS product, allowing people to sign up, and replacing the config/xml files with a nice web-based UI, and then charging something for it.