VASUDHA VISWAMURTHY

- Implemented distributed, multi threaded stress testing for AI model checkpoint encryption for largescale AI model training identifying multiple issues for model encryption at scale

- Improved reliability for AI Access Authorization service from 2 9s to 4 9s with regional deployments, client throttling, load testing and distributed configuration consumption methodologies.

- Established reliability standards for UX components for permission check services

- Improved permission check latency by 30x by implementing asset metadata retrieval through configerator delivery and consumption technology

- Set up User data privacy safeguards to uphold Meta's FTC and privacy commitment , automated evidence gathering for audit cycles

- Established client side SLOs , reliability standards for multiple tier 1 services in Meta

• Designed and implemented error categorization and classification for a compute engine library which improved reliability of the service by 2x.
• Improve visibility on the health of a library by modifying logging datasets, adding alerts and dashboards to inform failures within 10 minutes when occurred in production.
• Work with customers to design and track SLA for the service.

• Developed and enhanced features for setting up an automated integration testing environment for distributed compute and storage clusters for Facebook warehouse.
• Integrated the framework with continuous build of warehouse teams and reduced bad releases by 70%.
• Worked on turning up distributed storage clusters for warehouse use.

• Developed a web application framework for visualizing code coverage by test scripts and code dependencies in Cisco network operating systems (IOS), leading to improved code refactoring and code modularity.
• Designed and implemented an intelligent test-case selection and execution framework, resulting in 80% reduction in time taken for regression testing.
• Modelled and maintained a scalable graph database to store file and function hierarchies and dependencies, replacing CSV files and greatly reducing querying time.