Parth Shastri
CV
← All projects
Case studyBridgeweave

Equity baskets

Evolutionary search and computer vision for stock selection.

A stock market chart on a laptop screen
Photo by Markus Winkler / Unsplash ↗

At Bridgeweave, I worked on models that selected stocks for equity baskets. We combined evolutionary search with computer vision on price charts.

That gave us a lot of model candidates to work through. A large part of the job became testing them and making the next experiment easier to run.

A lot of backtesting#

We backtested over 1,000 models. The model iteration cycle came down from five days to one, which made a real difference to how quickly we could work through ideas.

01 / Model developmentFrom candidate models to evaluated baskets.
  1. Search & chart visionDevelop stock-selection models
  2. BacktestEvaluate model candidates
  3. IterateFive days → one day per cycle

A lower-risk basket#

I also built model switching around stock-level risk metrics for a lower-risk basket. This added another decision: which model to use given the risk information available.

The work involved both developing the models and building the evaluation process around them. With that many candidates, the speed of the experiment loop mattered.