Ilham Azhar

I’m a recent B.Sc. in Computer Science and Engineering graduate from BRAC University with a CGPA of 3.92, focused on AI/ML and vision-language modeling. In my undergraduate thesis, I built an open-weight pipeline for AI-driven context-aware trimming and segmentation of educational video content for focused learning.

I developed the system using Parakeet TDT 0.6B ASR, Qwen-3VL 4B (VLM), and Qwen3-4B-Instruct (LLM), then evaluated it on 20 human-annotated videos (4,672 segments). The model achieved 96.85% accuracy and 90.86% F1-score, outperforming the Gemini-3 baseline (81.63% accuracy) through dataset labeling, preprocessing, model evaluation, and error analysis.

I also gain practical product experience through my Urban Maid Service Platform (MERN-based system), where I designed API communication and supported deployment across Vercel and Railway. Alongside research, I served as a Student Tutor for Linear Algebra & Fourier Analysis, and I’m now seeking an AI/ML internship (and opportunities to teach and deepen my expertise in modern AI architectures).