Nazia Shehnaz Joynab

I'm a Software Engineer (DevOps) at Samsung R&D Institute Bangladesh. I graduated from the Dept of Computer Science and Engineering, MIST in May, 2023.

In my current position as a Software Engineer at Samsung Electronics, I am involved in the development and maintenance of an internal Content Delivery Network (CDN), named Artifact Delivery System (ADS). It is a distributed caching system designed to facilitate the rapid and efficient transmission of large binary artifacts across different global research centers of Samsung. The architectural framework of our system integrates proprietary network accelerators, smart caching mechanisms, proxy configurations, and a public cloud infrastructure to optimize artifact delivery for Samsung's employees and strategic partners. Within my present role, I am actively engaged in the development, deployment and monitoring of the system components using Django, Java, Bash, Fabric, MySQL, Docker and Quickbuild. Furthermore, I provide support to various research initiatives at SRBD. My team was amongst the finalists in an internal AI PoC competition where I significantly contributed to ML model development.

As an undergrad student at MIST, I participated in several programming contests and hackathons to develop my critical thinking, time management skills and problem-solving abilities. I conducted courses on programming language basics, object-oriented programming on Python as part of my extra-curricular activities.

Email  /  CV  /  Scholar  /  Linkedin  /  Github

Research

I'm interested in the application of ML in software engineering field as well as distributed ML. My current career goal is to pursue a PhD program in my field of interest.

We studied the efficacy of federated learning in cervical cancer prediction on three different experimental settings (2~IID, 1~Non-IID). Our proposed CNN-based FL architecture showed a test accuracy of 94.36% and 78.4% on an IID (Independent and Identically Distributed) and a non-IID setting respectively. In the mentioned IID setting, three clients were distributed images of all the five classes while in the non-IID setting, each client had images of one distinct class only. The results indicate that FL showed a significant performance over traditional ML algorithms.

Projects

• Laravel8
• Oracle
• PHP
• SQL

A web platform with features like role based authentication, appointment system, online Q/A forum, course addition/deletion and progress analysis system for special children.

• Flutter
• Cloud Firestore
• Machine Learning

Traditional ML models were trained to predict the risk of having cervical cancer based on user inputs of pre-defined questionnaire through the developed app. It was a department funded project.

• Deep Learning

Five pre-trained models were fine-tuned for recognizing 38 classes of Bengali Sign Language images. Using pre-trained models of Pytorch - Densenet121, VGG16, Mobilenet v3 small, Mobilenet v2, and Resnet50, we achieved test accuracy of 96.57%, 95.13%, 92.82%, 95.52%, 96.31% respectively.

Design and source code from Jon Barron's website