Founder & CEO, Smart SohayAI ResearcherDhaka, Bangladesh
AI researcher and founder building education technology for Bangladesh. My platforms have reached over a million learners.
Founder & CEO of Smart Sohay, data-driven exam preparation for SSC, HSC, admission and job exams. Before it I built Amar iSchool. My research in Bangla NLP and computer vision, 15+ peer-reviewed papers, informs the technology behind Sohay AI.
Figures as of Oct 2026
What I’m building
Smart Sohayস্মার্ট সহায়Founder & CEO · since 2025
Data-driven exam preparation for SSC, HSC, university admission and government job exams. It finds the topics a student is weak in, ranks what to study first for their target exam, keeps them practising until each topic is strong, then brings it back before it is forgotten.
Zone analysis, question bank analysis and spaced-repetition review work on one data layer. Sohay AI answers any topic by voice, camera or text, in Bangla and English.
1M+ expert-verified questions, tagged by exam, year, subject, chapter and topic
Google Play Console & internal analytics
1 Find the weak topic
Accuracy on every topic, from the student’s own answers.
2 Know what to study first
Not every weak topic needs to come first. For the target exam, the app marks the priority topics, and it isn’t just how often a topic appeared. With no weak topics yet, it still shows where to start.
3 Practise until it’s strong
Weak topics and exam priority combine into one daily pick: what to study today, and why. Progress is tracked from weak to strong.
4 Keep it strong
Our own modified SM-2 spaced-repetition schedule brings a topic back before it is forgotten, so a strong topic stays strong.
Steps 1–3 are real screens from a student’s account, shown in Bangla as students use them. Step 4 shows how the review schedule works.
How I think
At Amar iSchool I saw that video lessons alone don’t move exam scores. What helps is knowing exactly which topics are costing marks, then practising those until the score changes. Smart Sohay is built around that loop.
We spent a full year building the question bank before launch — 1M+ questions, every one reviewed by domain experts and mapped to subject, chapter and topic. That structure can’t be copied by scraping.
I work closely with teachers to understand what slows them down. In Bangladesh, video learning sells on trusted names. When the platform itself becomes a trusted brand, quality teachers can reach far more students through it.
Amar iSchoolSkill courses in Bangla, from research writing to programming and machine learning, also on GP Academy. AI/ML courses taught by engineers now at Meta and Google. The lessons from here shaped Smart Sohay. Still running; I stepped back from day-to-day operations to focus on Smart Sohay.
1M+ learners · Grameenphone Accelerator, Batch 6
amarischool.com
Pioneer AlphaR&D software company. Built Bangla NLP, computer vision and education products, and consulted for organisations including Tdh Bangladesh, DiagnoSmart and Hulkenstein. 10+ AI projects for local and international clients. Still running; I stepped back from day-to-day operations to focus on Smart Sohay.
BASIS National ICT Award 2020
pioneeralpha.com
BokBokNo-code AI agents for customer chat and social posting.
bokbok.appResearch
WACV Workshops (IEEE/CVF Winter Conference on Applications of Computer Vision)
@inproceedings{ashrafee2022realtime,
title = {Real-time Bangla License Plate Recognition System for Low Resource Video-based Applications},
author = {Ashrafee, Alif and Khan, Akib Mohammed and Irbaz, Mohammad Sabik and Al Nasim, MD Abdullah},
booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) Workshops},
year = {2022}
}IEEE ICMLA (International Conference on Machine Learning and Applications)
@inproceedings{khan2021endtoend,
title = {End-to-End Natural Language Understanding Pipeline for Bangla Conversational Agents},
author = {Khan, Fahim Shahriar and Al Mushabbir, Mueeze and Irbaz, Mohammad Sabik and Al Nasim, MD Abdullah},
booktitle = {2021 20th IEEE International Conference on Machine Learning and Applications (ICMLA)},
year = {2021}
}ICCIT (25th International Conference on Computer and Information Technology)
@inproceedings{alnasim2022brain,
title = {Brain Tumor Segmentation using Enhanced U-Net Model with Empirical Analysis},
author = {Al Nasim, MD Abdullah and Al Munem, Abdullah and Islam, Maksuda and Palash, Md Aminul Haque and Haque, MD. Mahim Anjum and Shah, Faisal Muhammad},
booktitle = {2022 25th International Conference on Computer and Information Technology (ICCIT)},
year = {2022},
organization = {IEEE}
}IC4IR (International Conference on Fourth Industrial Revolution and Beyond)
@inproceedings{palash2021bangla,
title = {Bangla Image Caption Generation through CNN-Transformer based Encoder-Decoder Network},
author = {Palash, Md Aminul Haque and Al Nasim, MD Abdullah and Saha, Sourav and Afrin, Faria and Mallik, Raisa and Samiappan, Sathishkumar},
booktitle = {International Conference on Fourth Industrial Revolution and Beyond (IC4IR)},
year = {2021}
}ICASERT (1st International Conference on Advances in Science, Engineering and Robotics Technology)
@inproceedings{hossain2019brain,
title = {Brain Tumor Detection Using Convolutional Neural Network},
author = {Hossain, Tonmoy and Shishir, Fairuz Shadmani and Ashraf, Mohsena and Al Nasim, MD Abdullah and Shah, Faisal Muhammad},
booktitle = {2019 1st International Conference on Advances in Science, Engineering and Robotics Technology (ICASERT)},
year = {2019},
organization = {IEEE}
}The Fourth Industrial Revolution and Beyond: Select Proceedings of IC4IR+
@inproceedings{palash2023incongruity,
title = {Incongruity Detection between Bangla News Headline and Body Content through Graph Neural Network},
author = {Palash, Md Aminul Haque and Khan, Akib and Islam, Kawsarul and Al Nasim, MD Abdullah and Shahjahan, Ryan Mohammad Bin},
booktitle = {The Fourth Industrial Revolution and Beyond: Select Proceedings of IC4IR+},
year = {2023},
publisher = {Springer}
}ICECIT (International Conference on Electronics, Communications and Information Technology)
@inproceedings{alnasim2021automated,
title = {An Automated Approach for the Recognition of Bengali License Plates},
author = {Al Nasim, MD Abdullah and Chowdhury, Atiqul Islam and Muna, Jannatun Naeem and Shah, Faisal Muhammad},
booktitle = {International Conference on Electronics, Communications and Information Technology (ICECIT)},
year = {2021},
organization = {IEEE}
}IEEE RAAICON (Robotics, Automation, Artificial-Intelligence and Internet-of-Things)
@inproceedings{palash2021finegrained,
title = {Fine-Grained Image Generation from Bangla Text Description using Attentional Generative Adversarial Network},
author = {Palash, Md Aminul Haque and Al Nasim, MD Abdullah and Dhali, Aditi and Afrin, Faria},
booktitle = {IEEE International Conference on Robotics, Automation, Artificial-Intelligence and Internet-of-Things (RAAICON)},
year = {2021}
}International Conference on Big Data, IoT, and Machine Learning
@inproceedings{irbaz2022realtime,
title = {Real-time Face Recognition System for Remote Employee Tracking},
author = {Irbaz, Mohammad Sabik and Al Nasim, MD Abdullah and Ferdous, Refat E},
booktitle = {Proceedings of the International Conference on Big Data, IoT, and Machine Learning},
year = {2022},
publisher = {Springer}
}HORA (International Congress on Human-Computer Interaction, Optimization and Robotic Applications)
@inproceedings{alnasim2020comparative,
title = {A Comparative Analysis on Bangla Handwritten Digit Recognition with Data Augmentation and Non-Augmentation Process},
author = {Al Nasim, MD Abdullah and Ferdous, Refat E and Pantho, Mahim Anzum Haque and Chowdhury, Atiqul Islam},
booktitle = {International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA)},
year = {2020},
organization = {IEEE}
}Data Driven Approaches on Medical Imaging, Springer, 2023. Author of 7 of the 10 chapters.
I have directly mentored 10+ students through their first research papers; several went on to MS and PhD programs abroad. I also mentored Amar iSchool’s Research Paper Writing course, taken by 10,000+ paid students.
Peer reviewer for IEEE Access and ICCIT.
Journey
Data-driven exam preparation built on topic-level zone analysis. 700K+ downloads in its first year.
Also set up telemedicine at the Tdh Rohingya camp, Teknaf, as IT specialist.
Founded Pioneer Alpha the same year, and built a Bangla customer-service chatbot as AI developer at Grameenphone.
Studied 2015–2019. Thesis supervisor: Faisal Muhammad Shah.
Machine learning intern at NybSys the same year (face recognition).
SSC, GPA 5.00.
Recognition
In photos
Books and teaching
Machine learning in Bangla, from Python libraries to deep learning, with code for every algorithm, interview questions and a research guide.
Buy on Rokomari
100 competitive programming and coding-interview problems, each with the explanation, the thinking strategy, the solution and C++ code. Written in Bangla with Md. Mahim Anjum Haque and Nirjoy Debnath; edited by Iftekhar Hakim.
Buy on RokomariContact
A 30-minute call is the fastest way to reach me. Pick a time on the calendar.
Book a 30-minute callTimes shown in your time zone
Open in Calendly