About Me
I am an Electrical and Electronics Engineer working at the intersection of Deep Learning and Computer Vision — focused on building efficient, scalable AI systems across the full research-to-deployment stack. My work spans knowledge distillation, object detection, and model compression, with an emphasis on production-grade applied research.
Beyond engineering, I am drawn to the Renaissance ideal of broad intellectual curiosity — pursuing ongoing study in Mathematics, Physics, Philosophy, and Music as disciplines that sharpen both analytical thinking and creative problem-solving.
Research Interests
- Knowledge Distillation & Model Compression — teacher-student training, PEFT, quantization
- Computer Vision — object detection, visual representation learning, DETR-family architectures
- Applied AI Research — closing the gap between research advances and production systems
Selected Projects
detrflow — End-to-End RT-DETR Object Detection Pipeline Production-ready object detection pipeline built on RT-DETR, achieving AP = 47.9 on COCO val2017. Includes a FastAPI inference API, HuggingFace Space deployment, and benchmark scripts. Accompanied by a peer-reviewed publication on knowledge distillation for RT-DETR (arXiv:2605.31191).
Student Capacity Moderates Knowledge Distillation Effectiveness Systematic study of Logit-KD and Feature-KD across three ResNet teacher-student pairs on CIFAR-10. Key finding: student capacity — not the teacher-student accuracy gap — is the primary moderating factor in KD effectiveness. Results reproduced across 3 seeds; interactive demo on HuggingFace Spaces.
Background
I hold a degree in Electrical and Electronics Engineering and have built a strong applied research foundation through academic work and hands-on projects — with a focus on taking models from research prototype to real-world deployment. I also maintain an active blog where I write about deep learning, research papers, and applied AI.
Get in Touch
Open to research discussions, collaborations, or exchanging ideas.