Md Romyull Islam
PhD Student in Computer Science
Kennesaw State University, GA, USA
I am a dedicated PhD student in Computer Science at Kennesaw State University, specializing in language model optimization and edge computing. With a strong academic foundation including a Master's degree from Austin Peay State University and a Bachelor's from Daffodil International University, I bring extensive research experience in machine learning and artificial intelligence.
My current research focuses on optimizing large language models for edge devices, contributing to the advancement of efficient AI deployment. I have published multiple peer-reviewed papers and actively contribute to the research community through conferences and collaborative projects. My technical expertise spans across edge computing platforms, and advanced ML techniques, making me well-suited for both academic research and industry innovation.
Research Interests & Ongoing Work
Language Model Optimization for Edge Devices
My primary research focus involves developing novel techniques for optimizing large language models to run efficiently on edge devices. This includes working with quantization methods, model compression, and hardware-aware optimization strategies to enable real-time inference on resource-constrained devices like Raspberry Pi and Jetson Orin.
Edge Computing and AI
I explore the intersection of artificial intelligence and edge computing, investigating how to deploy sophisticated AI models on devices with limited computational resources. My research contributes to making AI more accessible and practical for real-world applications where cloud connectivity may be limited or latency is critical.
Energy Efficiency in Small Language Models
Investigating the energy consumption patterns and efficiency opportunities in small language models deployed on edge devices. This research is crucial for sustainable AI deployment and extends the operational lifetime of battery-powered edge devices.
Adaptive Sparsity Driven Quantization-Aware Low-Rank Adaptation (ASQLoRA)
Currently developing a novel fine-tuning methodology that combines quantization techniques with Low-Rank Adaptation (LoRA) to create more efficient language models. This research aims to maintain model performance while significantly reducing computational requirements and memory footprint.
Publications
Publication Details
Professional Experience
Graduate Research Assistant
- Conduct research on language models and edge devices, focusing on optimizing models for edge hardware such as Raspberry Pi and Jetson Orin
- Fine-tune and optimize large language models, improving performance for real-time inference
- Collect, analyze, and interpret data from experiments to support research findings
NLP and Data Science Intern
- Scraped and processed data from PDF documents to build a robust knowledge base
- Assisted in developing a chatbot tailored for environment-friendly industries, using language models to enhance accuracy and response times
Graduate Teaching/Graduate Research Assistant
- Graded quizzes, homework, and proctored exams for machine learning and predictive analytics courses
- Assisted students in understanding machine learning concepts and statistical techniques for project work
- Applied advanced data visualization and querying techniques to analyze complex datasets
- Worked with environment and time series data, applying various machine learning algorithms to create optimal models
Machine Learning Intern
- Collaborated with the machine learning team to develop an AI-powered medical support system
- Labeled and preprocessed medical query data, achieving 86% model accuracy in automating query routing
- Applied multi-label classification techniques to ensure healthcare professionals' accurate query resolution
Technical Skills
Frameworks & Libraries
- TensorFlow
- PyTorch
- NumPy
- Pandas
- Scikit-Learn
- Matplotlib
- OpenCV
- NLTK
- Seaborn
- Keras
- PySpark
Platforms & Cloud
- Google Colab
- Amazon SageMaker
- Tableau
- Arduino
Machine Learning & AI
- Random Forest
- Decision Trees
- Gradient Descent
- Support Vector Machine
- Computer Vision
- Natural Language Processing
- Convolutional Neural Networks
- Recurrent Neural Networks
- Transformer
- Vision-Transformer
Language Models & Advanced AI
- Llama.cpp Inference
- LoRA Fine-tuning
- RAG (Retrieval-Augmented Generation)
- Quantization
- Edge Inference
Data Analytics & Statistics
- Regression Analysis
- Probability Theory
- Time Series Analysis
- Data Visualization
- Hypothesis Testing
- Data Augmentation
- Missing Value Handling
Database & Version Control
- SQL
- MySQL
- Git