Prof. Elakkiya Rajasekar
Associate Professor of Computer Science, BITS Pilani Dubai Campus
- Trustworthy AI
- Computer Vision
- Accessibility
The take
A benchmark score is a claim. Trust is earned in deployment.
AI systems are judged today by how well they score on curated test sets. Deployment is a different test. A model that performs well in the lab can be misled by small adversarial changes to its input, can retain personal data it should have forgotten, and can fail the users who were least represented in its training data.
Trustworthy AI means designing for those conditions from the start: robustness to attack, the ability to unlearn on request, and equal performance across every user group, including persons with hearing or speech impairment. My research on sign language technology has shown me that accessibility is the strictest test of all. If a system works there, it can be trusted elsewhere.
Editorial bio
Elakkiya Rajasekar works on the question of what makes AI dependable outside the lab: adversarial robustness, machine unlearning in language models, and multimodal systems for sign language recognition. She is associate professor of computer science at BITS Pilani Dubai Campus and associate head of the Anuradha and Prashanth Palakurthi School of AI (APPScAI).
Her book, Generative AI for Sign Language Recognition and Translation (IET), lays out end-to-end systems that serve persons with hearing or speech impairment – the strictest accessibility test she argues an AI model can face. She has led 20 funded projects, 16 as principal investigator, and has been on the Stanford–Elsevier top 2% world scientists list for three consecutive years. She is ACM-W Asia Pacific chapters chair and convened the first ACM and ACM-W professional chapters in the UAE.
Worth following if you want trustworthy-AI research that's tested where it matters, not where it benchmarks.
Key facts
- based in
- Dubai, UAE
- current
- Associate Professor, Department of Computer Science, BITS Pilani Dubai Campus · Associate Head, APPScAI · Associate Head, LEAD Satellite Centre
- prior
- Assistant Professor, School of Computing, SASTRA University (2018–2022)
- recognition
- Top 2% World Scientists, Stanford–Elsevier (2023-24, 2024-25, 2025-26) · 20 funded projects (16 as PI)
- community
- ACM-W Asia Pacific Chapters Chair · Convenor and Chair, Dubai ACM-W and ACM Professional Chapters
- focus
- Trustworthy AI · adversarial robustness · machine unlearning · multimodal sign-language systems
Featured work
Generative AI for Sign Language Recognition and Translation
End-to-end guide to AI-based sign-language systems: datasets, pose estimation, text-to-sign, video generation, responsible AI practice.
Trustworthy AI
Adversarial robustness for edge AI and machine unlearning in language models.
Sign Language Recognition and Generation
Multimodal models that translate between sign language and text.
ACM-W Asia Pacific Chapters Chair
Regional leadership for women in computing across ACM-W chapters.