Research

Peer reviewed

Publications

Author names in blue indicate my contribution. A complete and current list is on Google Scholar.

2026

Autism Action Dataset (AAD): A Comprehensive Video Benchmark for ASD Behavior Recognition Using Deep Learning

Shahriar, M. A., Uddin, M. Z., Rafi, M. R. I., Akter, M. F., Alnajjar, F., & Ahad, M. A. R.

IEEE Access, 2026.

First author Dataset & benchmark IEEE Xplore ↗

2025

Skin disease diagnosis using decision and feature level fusion of deep features

Zasim Uddin, M., Arif Shahriar, M., Schuller, B. W., Nadim Mahamood, M., & Atiqur Rahman Ahad, M.

Frontiers in Digital Health, 7, 1478688.

Medical imaging Feature fusion Frontiers ↗

2024

Deep learning with image-based autism spectrum disorder analysis: A systematic review

Uddin, M. Z., Shahriar, M. A., Mahamood, M. N., Alnajjar, F., Pramanik, M. I., & Ahad, M. A. R.

Engineering Applications of Artificial Intelligence, 127, 107185.

Systematic review ASD ScienceDirect ↗

2023

Autism Spectrum Disorder Classification via Local and Global Feature Representation of Facial Image

Mahamood, M. N., Uddin, M. Z., Shahriar, M. A., Alnajjar, F., & Ahad, M. A. R.

2023 IEEE International Conference on Systems, Man, and Cybernetics (SMC), pp. 1892–1897. IEEE.

Conference Facial analysis IEEE Xplore ↗

Code

Project	Built with	Code
Skin lesion classification with FLF & DLF Feature- and decision-level fusion for lesion classification.	Python, Vision Transformer, CNNs	GitHub ↗
Action recognition for ASD Early action recognition from video, alongside the AAD benchmark.	CNN-LSTM, 3D CNN, ViViT	GitHub ↗
OCR-based Bangla RAG chatbot Answers questions from Bangla PDF documents.	LangChain, Gemma 3, Surya OCR	GitHub ↗
Product competitor analysis Multi-agent framework extracting competitor insights.	CrewAI, LLaMA 3.2	GitHub ↗
Resume categorisation Categorises resumes by domain to streamline screening.	BERT, KNN, Decision Tree, Random Forest	GitHub ↗

Further experiments and notebooks: github.com/reyadhasan605 ↗

Applied research

Three years building machine learning systems that had to run in production, under latency and reliability constraints.

October 2025 – July 2026

TechnoNext Software Limited — Dhaka, Bangladesh

Foodi Recommendation System (live): led a heterogeneous graph neural recommender using GraphSAGE with multi-head attention, trained with a combined BPR and InfoNCE objective, reaching 25% HR@10. Deployed with gRPC, FAISS and Redis behind a real-time feature pipeline.
Foodi Order ETA Predictor (live): led a LightGBM order preparation-time predictor achieving 3.32 minute MAE with a statistical fallback path, and optimised gRPC serving for roughly 3× throughput.

August 2023 – September 2025

Next Solution Lab — Dhaka, Bangladesh

Virtual Try-On (VTON): developed a diffusion-based virtual try-on system with texture-preserving and pose-guided modules, improving realism and garment alignment while cutting inference time by 54% for cross-category garments.
Research Assistant Agent: designed and deployed an AI research assistant powered by LLaMA 3 to automate literature review and summarisation.