Interview questions were standard fare, the usual brain teasers; one guy was really obsessed with calculating covariance matrices. They had a wood-panelled library with all the scientific journals where I spent the time between interviews. The gentleman who was guiding me through the interview process told me how their brokers were constantly front-running their trades so a big issue for them was keeping their market impact minimal.
During the interview I offered some potential ideas for alpha: having done some research on these guys before the interview I knew they had experts on voice recognition so I suggested they could put a microphone in the trading floor and gauge sentiment via spectral analysis. They really liked the idea. Trading floors no longer exist so this alpha is long gone, if it ever worked at all!
But alas - their alpha is simply skill and hard work - there are no shenanigans or insider trading going on there. It's just time-series regression done perfectly. I bet at the time they were not using anything fancier than linear regression. But considering how large their data team was they probably had access to datasets (even today) which give them tremendous edge. Data is oil in this field.
The book "MindF*ck" [0] briefly describes a project within RenTec to build a statistical behavioural model of every individual citizen the USA, fed by enormous public and private databases. From this behaviour they could anticipate consumer patterns which could predict the markets way ahead of anyone else. The insane level of ambition in these projects gives you a clue as to the level of technology they potentially might have.