Scholar
Devleena Das
Google Scholar ID: xzt3VP0AAAAJ
Member of Technical Staff, AMD
Citations & Impact
Citations
514
H-index
10
i10-index
10
Publications
17
Co-authors
6
list available
Contact
Publications
Resume (English only)
Academic Achievements
- Published multiple papers including 'Forecasting LLM Inference Performance via Hardware-Agnostic Analytical Modeling' (Under Submission), 'DEFT-UCS: Data Efficient Fine-Tuning for Pre-Trained Language Models via Unsupervised Core-Set Selection for Text-Editing' (EMNLP 2024), 'CE-MRS: Contrastive Explanations for Multi-Robot Systems' (RA-L 2024), 'Interactive and Explainable Robot Learning: A Comprehensive Review' (Foundations and Trends in Robotics 2024), 'Reprogramming Pretrained Language Models for Antibody Sequence Infilling' (ICML 2023), 'State2Explanation: Concept-based Explanations to Benefit Agent Learning and User Understanding' (NeurIPS 2023), 'Subgoal-based Explanations for Unreliable Intelligent Decision Support Systems' (ACM IUI 2023), 'Explainable Activity Recognition for Smart Home Systems' (ACM TiiS 2023), and 'Explainable Knowledge Graph Embedding: Inference Reconciliation for Knowledge Inferences Supporting Robot Actions' (IEEE IROS 2022).
Research Experience
- Currently working at AMD on LLM optimizations and architecture enhancements. Previously held roles at Scale AI, Georgia Tech, Accenture Labs, Sony AI, and IBM Research, contributing to supervised fine-tuning of LLMs, agentic LLMs for academic advising, data-efficient fine-tuning, explainable RL, LLMs for scientific discovery, and multi-modal deep learning models for robotics and smart homes.
Education
- PhD in Artificial Intelligence, specific school information not provided.
Background
- An AI scientist and engineer bridging research and real-world deployment. Passionate about building AI systems that combine practical impact with new innovations, especially relating to efficiency, explainability, and human-alignment.
Miscellany
- Resides in the San Francisco Bay Area.
