Ph.D. Student, Civil Engineering
University of Alabama
Tuscaloosa, Alabama
I am a Ph.D. student in Civil Engineering at the University of Alabama, working in the Connected and Automated Mobility (CAM) Lab with Dr. Mizanur Rahman. My research focuses on machine learning, language models, and autonomous systems.
Before starting my Ph.D., I completed my B.Sc. in Software Engineering from the Department of Computer Science and Engineering, Islamic University of Technology. I worked as a Software Engineer (L2) at REVE Systems, developing state-of-the-art Bengali speech recognition and synthesis systems. I also co-founded MasterCourse, an education technology startup.
Machine Learning, Large Language Models, Autonomous Vehicles, Software Security, Natural Language Processing, Computer Vision, Indoor Navigation, Cybersecurity.
Developed a pipeline for crowdsourced and scraped Bengali speech data at REVE Systems. Fine-tuned Wav2Vec2 for speech recognition and VITS for speech synthesis, and built public validation and serving systems. Used noisy audio during training to improve robustness and evaluated synthesized audio with multiple quality metrics.
Studied video anomaly detection using joint positions and graph convolution networks. Proposed DD-Net (Double-feature Double-motion Network), a lightweight architecture that reduced prediction time without compromising accuracy.
Advances in Transportation Cybersecurity and Resiliency, pp. 97–146. World Scientific, 2027. [Preprint]
arXiv preprint arXiv:2609.09386, 2026.
Accepted for TRB 2027 Presentation.
arXiv preprint arXiv:2608.13450, 2026.
Accepted for TRB 2027 Presentation.
Transportation Research Record, 2026. [Preprint]
arXiv preprint arXiv:2508.08120, 2025.
arXiv preprint arXiv:2508.08100, 2025.
Accepted for TRB 2026 Presentation and IEEE Journal of Indoor and Seamless Positioning and Navigation.
Built a backend for speech recognition and synthesis with Python FastAPI and Redis. Services communicate through message brokers and can be scaled independently.
Built a telephone-integrated voice bot that processes 20 ms audio chunks, converts voice commands to prompts, and returns speech responses.
Deployed a PyTorch TorchServe inference pipeline with Docker. Converted model checkpoints to ONNX for faster inference with ONNX Runtime.
Developed a real-time application that translates American Sign Language video into speech and English speech into ASL using prerecorded video.
Outside of research, I share interests in playing video games (mostly VALORANT) and self-hosting free (as GNU intended) software.