About Me

I am an Assistant Computer Programmer at Shahjalal University of Science and Technology (SUST), Sylhet, Bangladesh, where I contribute to the development, operation, and maintenance of the university’s core software systems and digital infrastructure.

My responsibilities include developing and maintaining major university platforms, including the official website, student service portal, and result processing system. I also manage DevOps operations for the undergraduate online admission system, covering application deployment, load balancing, monitoring, infrastructure management, and task automation.

Before moving into my current role, I served as an Assistant Network Engineer at SUST, where I worked on the university’s core network and critical ICT infrastructure. This combination of software engineering, DevOps, and network engineering has given me hands-on experience across the full lifecycle of large-scale institutional systems.

Education

I received my Bachelor of Science in Computer Science and Engineering from Shahjalal University of Science and Technology (SUST), Bangladesh, with distinction.

Professional Journey

I began my professional career in 2020 as a Junior Software Engineer at REVE Systems in Dhaka. There, I worked on an enterprise chatbot and contributed to improving its accuracy by approximately 5% through entity extraction, intent matching, and contextual analysis.

In 2021, I joined the Child Health Research Foundation (CHRF) in Dhaka as a Bioinformatician. My work focused on analyzing large-scale genomic datasets generated by the organization’s sequencing laboratory and developing computational workflows to extract meaningful biological insights.

My professional journey has therefore spanned software engineering, natural language processing, bioinformatics, network engineering, and DevOps, while my research has primarily focused on spoken language processing, natural language processing, and computational biology.

Research

Spoken Language Processing

My research journey began in 2018, during my undergraduate thesis, when I joined the Bangla Speech Processing Lab at SUST under the supervision of Prof. Mohammad Shahidur Rahman. My initial research focused on developing natural-sounding text-to-speech (TTS) systems for Bangla, a low-resource language.

Working with an undergraduate peer, I developed a neural network-based Bangla TTS system trained on approximately 30 hours of single-speaker speech data prepared in our laboratory. The system predicts acoustic features from linguistic inputs using neural networks and generates speech waveforms through a vocoder. It achieved a Mean Opinion Score (MOS) of 3.75/5 and resulted in my first-author conference publication at ICBSLP 2019.

We subsequently developed an end-to-end Bangla TTS system capable of synthesizing speech directly from Bangla text. The system achieved a MOS of 3.79/5 and led to a publication at ICSCT 2021, where I was the second author.

After completing my undergraduate degree, I continued collaborating with researchers in the lab to improve the naturalness of Bangla synthetic speech. We developed a TTS system combining a Transformer-based acoustic model with a neural vocoder, achieving a MOS of 4.6/5. The system received national recognition by securing first place in the “AI for Bangla 2.0” competition.

Natural Language Processing

More recently, my research has focused on Bangla text normalization, an essential preprocessing component of text-to-speech systems. Text normalization converts non-standard written expressions—such as numbers, dates, abbreviations, and symbols—into the forms in which they are naturally spoken. Because this process is highly context-dependent, errors in text normalization can lead to incorrect pronunciation even in otherwise high-quality TTS systems.

To address this problem, we developed a hybrid LLM-augmented Bangla text normalization framework that combines a large language model for identifying the semiotic classes of non-standard words with rule-based modules for verbalization.

As part of this work, we created and publicly released a Bangla text normalization dataset, supporting resources, and source code to facilitate further research. The work resulted in my first-author publication in the Natural Language Processing Journal.

Computational Biology

During the COVID-19 pandemic, I joined Omics Lab, an interdisciplinary research group, as a remote research collaborator. My first project investigated how the choice of genome assembler affects the quality of de novo SARS-CoV-2 genome assembly and downstream variant calling.

I designed the experimental workflow and developed the computational programs required to systematically benchmark different assemblers. The study provided practical recommendations for SARS-CoV-2 genomic analysis and resulted in a peer-reviewed publication in Briefings in Bioinformatics, where I was the second author.

Following this work, I led the development of VirusTaxo, a computational tool for taxonomic classification of viruses from genome sequences. The system was evaluated across 402 DNA viral genera and 280 RNA viral genera. I designed the core algorithm, implemented the complete software pipeline, and conducted the experimental evaluation. This work resulted in my first-author publication in Genomics.