Scholar
Sai Ashish Somayajula
Google Scholar ID: NU1fOrwAAAAJ
Oracle AI; University of California, San Diego; Indian Institute of Technology, Hyderabad.
Citations & Impact
Citations
125
H-index
5
i10-index
4
Publications
19
Co-authors
12
list available
Publications
Resume (English only)
Academic Achievements
- Paper 'Token-Specific Watermarking with Enhanced Detectability and Semantic Coherence for Large Language Models' accepted to ICML'24; 'Transformer Architecture Search for Improving Out-of-Domain Generalization in Machine Translation' accepted to TMLR'24; Work on continual pretraining accepted to TMLR'25; DreamPRM method achieved first place on the MathVista leaderboard; Research on re-weighting pretraining objectives for task-adaptive pre-training accepted to TMLR'25; Work on adaptive data augmentation for long COVID-19 literature classification accepted to Scientific Reports, Nature Portfolio; Paper 'Improving image classification of gastrointestinal endoscopy using curriculum self-supervised learning' also accepted to Scientific Reports, Nature Portfolio.
Research Experience
- Research Internship at Apple Park, Cupertino, with the Input Experience NLP team, working with industry experts Dr. Vivek Kumar Rangarajan, Leo Xu, Shivangi Mahto, and Barada Acharya; Teaching Assistant for ECE271A Statistical Learning course, collaborating with Professor Nuno Vasconcelos; Gave a talk at the IEEE YP Graduate Seminar Series - Session 4.
Education
- PhD: Department of Electrical and Computer Engineering, University of California, San Diego, advised by Professor Pengtao Xie; Bachelor's: Indian Institute of Technology, Hyderabad, major in Electrical Engineering (minor in Computer Science and Engineering), advised by Professor Sumohana S. Channappayya and Professor Adity Siripuram, secured the second-highest CGPA in the B.Tech program across all departments, and was twice awarded the academic excellence award.
Background
- Research Interests: Optimizing language models for specialized domains, particularly healthcare. Long-term vision is to leverage these models for scientific discoveries. Inspired by the COVID-19 pandemic, focusing on developing efficiently trained language models to rapidly analyze data and provide valuable insights for vaccine development, etc.
Miscellany
- Feel free to reach out via email, LinkedIn, or GitHub to discuss ideas.
