Scholar
Ravi Ghadia
Google Scholar ID: PWJstCUAAAAJ
UT Austin
Citations & Impact
Citations
4
H-index
1
i10-index
0
Publications
2
Co-authors
7
list available
Publications
Resume (English only)
Academic Achievements
- Involved in multiple projects including but not limited to: Research work at the University of Texas at Austin, GPU architecture optimization at NVIDIA, and RL-related research at H2Lab, University of Washington.
Research Experience
- Currently a research assistant at the University of Texas at Austin, focusing on improving inter-token generation latency during LLM inferencing by tuning KV cache size using predictive ML methods. Previously worked as a GPU Architect at NVIDIA, Bengaluru, where he optimized GPU Architecture, built ML-powered predictive models for future product performance, and helped debug inefficiencies in current products, especially for the gaming market. Additionally, worked as a part-time researcher at H2Lab, University of Washington, with Prof. Prithviraj Ammanabrolu, deriving rewards from freeform linguistic feedback to train Large Language Models using RL.
Education
- Master's: Expected to graduate from the University of Texas at Austin in ECE by 2025, advisor: Dr. Poulami Das; Bachelor's: Graduated from IIT Kharagpur in Electronics and Electrical Communication with a minor in Computer Science in 2021, thesis: 'Training Generative Dialog Models using Reinforcement Learning with Learned Reward Systems', advisor: Dr. Pawan Goyal.
Background
- Research interests: Designing efficient systems for running massive AI models, using statistical methods, reinforcement learning, and other optimization algorithms to design efficient end-to-end AI systems; particularly interested in NLP (especially Large Language Models), RLHF, and interactive systems.
Miscellany
- Hobbies include trekking, calisthenics, cooking (traditional Indian cuisines), and playing guitar. Also enjoys reading blogs about tech-stuff and human evolution.
