Dr. Janibul Bashir
Assistant Professor
National Institute of Technology Srinagar
About Me
I am Dr. Janibul Bashir, a researcher dedicated to advancing the frontiers of Artificial Intelligence (AI) and Computer Architecture. I lead Gaash, a research group at NIT Srinagar focused on designing and optimizing innovative computer architectures that push the boundaries of AI applications. Our mission is to develop efficient, scalable, and adaptable systems that meet the dynamic demands of modern AI.
At Gaash, we are pioneering research in several cutting-edge areas:
AI-Driven Computer Architectures: We focus on creating high-performance architectures specifically optimized for complex AI applications, with an emphasis on achieving new levels of efficiency and adaptability.
Federated Learning and Transformers: By fusing federated learning with transformer architectures, we are exploring ways to enhance privacy-preserving AI that generalizes effectively across diverse datasets while safeguarding user data.
Transformers for Large Language Models (LLMs): Our work with transformers extends to large language models, with a particular emphasis on zero-shot and in-context learning, enabling these models to perform well on new tasks with minimal or no additional training.
Vision Models: CNN and Transformer Synergies: In computer vision, we are combining convolutional neural networks (CNNs) with transformers to develop hybrid models that surpass current benchmarks in image recognition and classification.
Events and Accomplishments
[July 2024] Patent Granted on Optical NoCs for GPUs
[Jan 2024] Research project accepted for funding by JKST&IC as Co-PI. [7.5 Lakhs]
[Jan 2024] Research project accepted for funding by JKST&IC as PI. [8 Lakhs]
[Jan 2022]: Research project accepted for funding under SERB SRG scheme as PI - Completed [27 Lakhs]
[Oct 2021]: Paper "FreqCounter: Efficient cacheability of encryption and integrity tree counters in secure processors." accepted by Journal of System Architecture.
[June 2021]: Paper "Dynamic MTU: A smaller path MTU size technique to reduce packet drops in IPv6" accepted by Journal of Computer and Information Sciences.