Efficient and Distributed Deep Learning Systems
Distributed training, optimization, and deployment of large-scale models, from multi-GPU cluster scheduling to model compression.
Virginia, USA
Sultanul Ovi
MS Student in Computer Science @George Mason University
PhD Applicant, Spring / Fall 2027Seeking Summer 2027 Internship
My work sits between making deep learning systems efficient enough to run and reliable enough to trust. On one side that is distributed training and compression for large models. On the other it is knowing when a model in a hospital or a charging station should not be believed.
I am finishing my MS in Computer Science (Machine Learning Concentration) at George Mason University, with 6+ years in research, 5+ in teaching, and 2+ across leadership, volunteer, and industry roles. I work with the AI in Cyber (AIC) Research Lab on the security of EV charging infrastructure.
Research Focus
LLM Agents for Cybersecurity and Cyber-Physical Systems
Agentic approaches to attack detection & vulnerability analysis in systems like EV charging infrastructure & IoT devices.
Trustworthy AI for Sensitive Applications
Models that can be trusted in healthcare, mental health, and accessibility, where a wrong prediction has a real cost.
Recent Updates
Appointed as a GTA for IT 223: Information Security Fundamentals, Fall 2026
Aug 2026
Reviewer for two Majestic journals, AI Innovation and Information Technology
Aug 2026
Selected Research
SNAM 2026
A Dual Pipeline Machine Learning Framework for Automated Multi Class Sleep Disorder Screening Using Hybrid Resampling and Ensemble Learning
Social Network Analysis and Mining, Springer Nature, 2026
Q1
Journal
NCAA 2026
Performance Characterization of Distributed Deep Learning Strategies: A Quantitative Evaluation of DDP, FSDP, and Parameter Server Architectures on GPU Cluster
Neural Computing and Applications, Springer Nature, 2026
Q1
Journal
ICDM 2025
Quantitative Currency Evaluation in Low-Resource Settings through Pattern Analysis to Assist Visually Impaired Users
25th International Conference on Data Mining (ICDM), Workshop: Data Mining for Reliable Decision Making, IEEE, 2025
A*
Conference
(Workshop)
Featured Projects
CompletedMath2LaTeX: Equation OCR to LaTeX with Qwen2 VL
A vision language app that fine tunes Qwen2 VL with LoRA to convert equation images into LaTeX, with accuracy evaluation and a Gradio demo.
PyTorchLoRAUnslothGradioQwen2-VL
CompletedNaturalQuery: AI Powered SQL Generator
Transforms natural language into optimized SQL queries with AI driven intelligence.
GPTFastAPIMySQLStreamlit
OngoingFER for Autistic Children
Deep learning based early emotion recognition in children with autism: dataset insights, model enhancements, and design strategies.
Facial ExpressionPyTorchAutism
Technical Skills
Programming Languages
PythonJavaGolangC/C++JavaScriptSQLBash
Data Science & Analytics
PandasNumPyMatplotlibSeabornTableauPower BI
Machine Learning & Deep Learning
PyTorchTensorFlowKerasHugging Face TransformersScikit-learn
LLMs & Advanced AI
LangChainAgentic AI frameworksRAGTransformer Fine-Tuning (LoRA)Knowledge DistillationModel Compression (Quantization, Pruning)LLM Evaluation
Security & Networking
OCPPISO 15118OCPIOpenADRCPS/OT SecurityNmapMetasploitWiresharkNessusMininetRYU Controller
Distributed Computing
Multi-GPU TrainingHPCMPISlurmRayDeepSpeed
Tools & Platforms
AWSGCPAzureLinuxDockerKubernetesGitMLflow
Application Development
FastAPIStreamlitGradioChromaDBNodeJSExpressJSMongoDB
Hobbies
Apart from research, I enjoy exploring new technologies, solving problems through competitive programming, and building tools that fix problems I keep running into. I also try to give back to the community by engaging in different volunteer opportunities. In my free time, I like playing chess and watching movies and TV series. I enjoy collecting and preserving currencies from across the world.