Buy-Side Quantitative Research

The problem

A research idea at an asset manager is worthless until it survives contact with the data, the backtest, the portfolio manager, and finally production. Most of them do not, and the expensive failure is the one that looks good in a backtest because of a defect in how the data was constructed.

Approach

1. 01

   Constructed and cleaned the data before modeling it, since point-in-time correctness and survivorship handling decide whether a result means anything.

2. 02

   Built signal and valuation models across global equities and fixed income, working directly with the portfolio managers rather than delivering research over a wall.

3. 03

   Applied time-series and spectral methods carried over from electrical engineering, where separating signal from noise in a nonstationary series is the entire discipline.

4. 04

   Wrote the production code that ran the accepted results, so research and implementation were not two separate jobs.

Stack