Scale Events
+00:00 GMT
Sign in or Join the community to continue

Using ML to Improve Labeling Quality on Scale Rapid

Posted Apr 07, 2022 | Views 3.2K
# Tech Talk
Share
speakers
avatar
Zhichun Li
Head of Rapid @ Scale AI

Zhichun Li is the Head of Scale Rapid at Scale AI. She built the team from scratch with a focus on providing the fastest way to production-level quality labels within a day, with no data minimums. As an early employee of the company, she built up the infrastructure for our supply ops system and scaled up our 3D Sensor Fusion product. Prior to Scale AI, Zhichun studied CS at CMU and earned her MBA from Yale SOM.

+ Read More
avatar
Bryan Chen
Lead Engineer, Rapid @ Scale AI

Bryan leads quality efforts at Rapid, ensuring that any use case involving images, videos, text, and more can be supported on Rapid to quickly obtain high quality annotations. Prior to Scale AI, Bryan studied computer science and mathematics at MIT, and was a top competitor in national algorithmic programming contests such as the USACO.

+ Read More
avatar
Kiko Ilagan
ML Lead, Rapid @ Scale AI

Kiko works on developing and applying machine learning driven systems within Rapid to improve measurement, annotation quality, and overall efficiency. Before joining Scale, he was an ML engineer at Meta (formerly Facebook) focused on personalization and recommendation models for the Facebook Jobs team.

+ Read More
SUMMARY

Scale Rapid helps ML teams iterate quickly with an on-demand, end-to-end, labeling solution with rapid turnaround times. In this tech talk, we discuss the challenges to build quality infrastructure for crowd management, then discuss the role machine learning can play in:

  • Achieving high quality even with subjective tasks
  • Flagging potential problems with evaluation tasks
  • Suggesting improvements for project health

We also discuss where ML can accelerate and scale human insights, the challenges of creating generalizable quality mechanisms, and the trade-offs that must be considered between accuracy and generalizability for such ML solutions.

+ Read More

Watch More

28:45
How Scale Uses ML to Improve Label Quality and Labeler Productivity
Posted Oct 06, 2021 | Views 3.1K
# TransformX 2021
# Breakout Session