Description

Engineers spend a lot of time during on call to fix operational Spark failures and retrying jobs or fixing upstream data quality issues, using an Agent to automate this loop can save a lot of time and effort. This session will explore how we can use self-correcting agents to provide details on data quality and handle operational failures, this would reinstate the trust in the dataset the company owns to make business decisions and make downstream executive dashboarding show the correct/trusted results. This talk also plans to cover common design patterns, human-in-loop workflows, guardrails for using agents in production, MAS (multi agent systems) and so on.

Engineers  spend a lot of time during on call to fix operational Spark failures and  retrying jobs or fixing upstream data quality issues, using an Agent to  automate this loop can save a lot of time and effort. This session will  explore how we can use self-correcting agents to provide details on data  quality and handle operational failures, this would reinstate the trust in  the dataset the company owns to make business decisions and make downstream  executive dashboarding show the correct/trusted results. This talk also plans  to cover common design patterns, human-in-loop workflows, guardrails for  using agents in production, MAS (multi agent systems) and so on.

Details

October 29, 2026

2:45 pm

-

3:30 pm

Northpoint Ballroom 2

Add to Calendar

Track:

AI & ML

Level:

Intermediate

Tags

No items found.

Presenters

Swetha Sekhar
Staff Data Engineer
OpenAI

Swetha is a Staff Data Engineer in Safety Systems, focused on building the data foundations that help teams measure, evaluate, and improve model safety. Her work spans safety measurement, ground truth data pipelines, and classifier evaluation infrastructure, with an emphasis on making complex safety signals reliable, interpretable, and usable at scale.By combining deep data engineering expertise with strong systems thinking, she helps ensure that safety evaluations are grounded in high-quality data, reproducible pipelines, and trustworthy metrics.Beyond her technical role, Swetha is an active community contributor and a strong advocate for knowledge sharing. She serves as an editor for a Substack newsletter community and regularly engages through speaking, writing, and community-building efforts. Across both her engineering and community work, she brings a thoughtful, collaborative approach to helping people reason more clearly about complex systems, data, and impact.