Quantitative analysis

You build machine learning models that forecast market moves and put them into live trading at a firm that trades its own money. DRW trades with its own capital, so nothing you build is sold to a client. A model either makes money on the firm's book or it does not. The seat is a machine learning researcher. You take methods such as transformer based architectures, the kind of model behind modern language tools, and apply them to forecasting problems in markets.

Forecasting here means predicting something about a price, a volume or a flow a little ahead of time, well enough that a trading strategy can act on it. The job covers the whole chain, from raw data through to a model running in production, which means live against real markets rather than in a test.