Yasoda Lasiru

Evaluated and tracked algorithm performance of Vision Language Models (VLMs) under data poisoning attacks across multiple model architectures and maintained detailed experimental records and benchmarks. Implemented and benchmarked existing defense techniques to mitigate adversarial data poisoning and produced technical documentation of findings and results. Designed and implemented a defense mechanism against ShadowCast data poisoning attacks, validated experimentally across multiple architectures, contributing to a co-first authored paper currently under review at TMLR.

Intern - AI Engineering - Temasek Laboratories, Singapore University of Technology and Design - Singapore

(2024-12 - 2025-06)

• Evaluated and tracked algorithm performance of Vision Language Models (VLMs) under data poisoning attacks across multiple model architectures and maintained detailed experimental records and benchmarks.
• Implemented and benchmarked existing defense techniques to mitigate adversarial data poisoning and produced technical documentation of findings and results.
• Designed and implemented a defense mechanism against ShadowCast data poisoning attacks, validated experimentally across multiple architectures, contributing to a co-first authored paper currently under review at TMLR.

Bachelor of the Science of Engineering (Hons.) - Electronic and Telecommunication Engineering - University of Moratuwa (2022-03 - 2026-06)

G.C.E. Advanced Level Examination - Physical Science - Sanghamiththa Girls' College (2007 - 2020)