Culture / Development / Machine Learning

Glassdoor: Don’t Sacrifice Performance for New Features

14 Dec 2020 3:00pm, by

Welcome to The New Stack Makers: Scaling New Heights, a series of interviews with engineering managers who talk about the problems they have faced and the resolutions they sought, conducted by guest host Scalyr CEO Christine Heckart.

Bhawna Singh had two mandates at Glassdoor when she started as chief technology officer and senior vice president of engineering: open an office in San Francisco to access the region’s talent pool, and rebuild the search vertical for job results. Glassdoor is a job and recruiting site that offers services that allow people to see information such as company reviews, salary reviews, and benefits that a potential employer offers.


Scaling New Heights EP #7 – Glassdoor: Performance Matters

Also available on Apple Podcasts, Google Podcasts, Overcast, PlayerFM, Pocket Casts, Spotify, Stitcher, TuneIn

To improve the quality of search, Heckart had to set metrics that the team would trust. Performance challenges surfaced when the team focused its efforts on the tactical aspects of architecting the platform. They had tuned the system for quality; building out the deployment infrastructure and adding machine learning models. But the work made the system heavier and less performant.

By this time, the site was getting more traffic and the performance issues had worsened, making it more difficult to scale. Quality improvements were pulled back to focus on performance — necessary because the team doing the work was already pretty small.

The team then built load and performance testing features into their deployment pipeline, to make sure they were comparing the performance of every release. Plus they moved back to their original roadmap of quality improvements.

The work paid off. Performance did get better with the infrastructure and work moved on to the next adventure: moving off the monolith.

A newsletter digest of the week’s most important stories & analyses.