The Research Wing mentors DIU students through literature reviews, experimental research, and submissions to top-tier venues in ML, HCI, systems, and theory.
Development wing
Faculty-guided research groups with consistent supervision and feedback.
Weekly reading circles covering landmark and latest papers — from attention mechanisms to distributed consensus.
Supervised research tracks in ML, NLP, computer vision, and theoretical CS, culminating in publishable work.
Guidance on writing, peer review, poster prep, and submission to IEEE, ACM, and Bangladeshi venues.
Every wing operates on a semester-long curriculum with weekly sessions, office hours, and progression checkpoints.
Research cycle
Research progress is structured with clear milestones, reproducibility requirements, and supervised writing loops.
Week 1
Identify a meaningful research question by surveying prior work, gaps, and practical impact.
Week 2-4
Define hypotheses, baselines, datasets, and reproducible methodology before model training.
Week 5-6
Run ablations, perform statistical checks, and validate claims with clear visual evidence.
Week 7-8
Draft the manuscript, iterate with peer review, and prepare for conference or journal submission.
Members join domain labs while sharing methods, tooling, and review practices across the research wing.
Model development, benchmarking, and robust evaluation for applied and theoretical ML studies.
Language modeling, retrieval workflows, evaluation design, and multilingual research directions.
Computer vision and interaction research with user studies and human-centered experimentation.
Algorithmic rigor, systems performance studies, and mathematically grounded CS research.
Every project is expected to be reproducible, evidence-backed, and ready for academic scrutiny.
Outcome focus
Publish with confidence, not guesswork.
Members conclude each cycle with documented experiments, manuscript drafts, and venue-ready submission plans.
Join the wing
Apply when recruitment opens, join a supervised lab, and build a publication-oriented portfolio step by step.
