I’m Abrar Faiaz — a researcher and engineer working at the intersection of scientific machine learning, multimodal AI, and decision-support systems. My work spans cyclone forecasting, geospatial intelligence, AI-powered search, and LLM systems for real-world impact.
I’m a Civil Engineering graduate from BUET, researching machine-learning methods for complex environmental and physical systems. My work explores physics-informed learning, representation learning, and multimodal AI for forecasting, retrieval, and decision support in high-impact real-world settings.
B.Sc. Civil Engineering
Bangladesh University of Engineering & Technology (BUET) · 2021–2026
Notre Dame College, Dhaka — HSC
Leading development of an AI-powered search engine for wearables spanning 400+ stores across Bangladesh. Built the end-to-end search & discovery stack — embedding-based retrieval, image search, and scalable scraping pipelines for product matching, ranking, and search across fragmented e-commerce sources.
Developing an AI-powered cyclone-prediction system for the Bangladesh Meteorological Department — combining ML forecasting of cyclone metrics with a retrieval-augmented LLM that generates localized advisories for actionable decision support in high-impact climate settings.
01
Abrar Faiaz, Alphy Shaharin, Md. Mafizur Rahman, Md. Bazlur Rashid, S.M. Quamrul Hasan
DOI: 10.1016/j.eswa.2026.131972 ↗
02
Abrar Faiaz, Rajit Raihan, Md. Mafizur Rahman, Md. Bazlur Rashid, S.M. Quamrul Hasan
03
Rafi Uddin, Abrar Faiaz, Sk. Rakibul Islam
DOI: 10.1007/978-981-97-0751-5_19 ↗
04
Abrar Faiaz
Abrar Faiaz, Dipankar Mitra, Ranat Das Prangon
DOI: 10.1145/3704522.3704541 ↗Abrar Faiaz, Nowshin Nawar
DOI: 10.1145/3704522.3704549 ↗Civil Engineering Conference in the Asian Region (CECAR10)
Software that predicts how your audience will feel — modeling emotional response to creative and content before it ships.
Visit the dashboard ↗Transformer NLP models, LLM & LangChain pipelines, PyTorch & TensorFlow.
huggingface.co/Abrar20 ↗ OngoingDeep learning, LLMs, VLM & geospatial analysis for Forewarn Bangladesh & BMD.
Learn more ↗ Innovation · HardwareSolving the 4-million-riders problem — ANSYS-simulated HEPA filtration with sensor-driven brushless-DC airflow, prototyped through 3D printing.
View the design ↗
