Hey there!!! I am Samiul.

I am one of the co-founders of bengali.ai. My job here is to mostly extract and clean raw data for making datasets. I handle outliers and missing data and analyze input correlations and variance to make sure the data can be easily used to train models.

I am deeply passionate about machine learning and regularly try and find novel applications. As such, I have worked in a range of  areas from music information retrieval to steganography. If you have any questions about my research or have any relevant ideas in mind, feel free to e-mail me at samiul.a@bengal.ai.

I graduated from Bangladesh University of Engineering and Technology in October, 2017. I currently work at Samsung Research and Development Institute as an iOS developer.

What's New


My research interests include machine learning, optimization, computer vision and robotics. My motivation for studying these stems from my love of mathematics and physics. I am deeply convicted to exploring the possibilities of computer intelligence and solving the mysteries of human cognition. Personally, being fond of physics and math, I usually try to explain phenomenon under physical laws and equations. Interestingly, the workings of the mind have managed to elude my perception. So, it is my wish to understand and somehow emulate true creativity.

  • Parkinson’s Disease Classification

    I along with a colleague of mine in Samsung Research collaborated with Professor Dr. Taufiq Hasan on Parkinson’s disease diagnosis by using spiral drawings made by patients by a stylus on a graphics tablet. Since the available dataset is small, we are currently in the process of collecting more data for the project. If accurate enough, the project can provide a way for people in remote areas to have access to preliminary diagnosis of Parkinson’s Disease.

  • Real Time Beat Tracker

    This was done as part of IEEE Signal Processing Cup 2017. The goal was to design a real time musical beat tracker on an embedded platform and demonstrate creative ways to represent beats. I worked first on a three-layer LSTM network based model and then a Viterbi algorithm based model. But both of this were too tasking for the raspberry pi. The team finally went for a simple dynamic programming implementation that stored previously computed values to save compute power. It was able to work in real time with slight fall of accuracy. Our work won honorable mention in the final round of the competition and we later published our work at IEEE Multimedia Information Processing and Retrieval Conference 2018.

  • Control System of High density DC-DC Converter

    I worked on the control system of a DC-Dc converter when I participated in IEEE future Energy Challenge. This was also a research competition. Here, we had to build an extremely efficient DC/DC converter with a very small form factor. I worked on the control system of the converter. The switching circuits had to work at high frequency and had to adjust for voltage and load fluctuations. I used my previous experience working with viterbi algorithm to implement accurate switching inside the converter. Our work was accepted into the second round of the competition and we presented our progress at Applied Power Electronics Conference 2017. The project became too costly for us to continue our work properly and we could not progress further but it proved a great learning experience for me.


  • I. A. Hussaini, A. I. Humayun, S. I. Foysal, S. Alam, A. Masud, A. Mahmud, R. Islam, N. Ibtehaz, S. U. Zaman, R. Hyder, S. S. Chowdhury, and M. A. Haque

    Predictive Real-time Beat Tracking from Music for Embedded Application

  • S. Alam, T. Reasat, R. Mohammad Doha, A. I. Humayun

    NumtaDB – Assembled Bengali Handwritten Digits

Awards And Honors


MARCH 5, 2017

Designed a real time musical beat tracking system


MARCH 26, 2017

Designed, simulated and prototyped a high efficiency, high density DC/DC converter


NOVEMBER 11, 2016

Built a 3D- scanner with the purpose of enabling young innovators to learn 3D design and printing.


MAY 12, 2015

Built a maze solving robot that can go through specific points in the maze while avoiding obstacles.



Awarded the scholarship to fund education for 8 semesters.