Who am I ?

Officially tagged as Fahimul Hoque Shubho. I am currently pursuing my Master's degree in Computer Science and Engineering, specializing in Intelligent Systems, at North South University. Also working as a Machine Learning Engineer, developing and deploying impactful real-world ML and AI solutions globally.

What i'm doing nowadays

Research

Computer Vision, Zero-shot Learning
Artificial Intelligence

Machine Learning, Deep Learning, Agentic AI
Automation

Automation of variety of tasks using Python like a Ninja
Gaming

Valorant, Anno 1800, Euro Truck Simulator 2

Find me on

Resume

Experience

Machine Learning Engineer
November 2025 — Present
Ajentica
- Working on leveraging Agentic AI for delivering personalized Behavioral Health services.
- Developing Multi-Agent orchestration for implementing efficient RAG into existing systems.
- Utilizing LLM to enrich content metadata for optimized search and filtering for a Search Engine.
- Building stateful Multi-Agent systems with shared memory and collaborative reasoning.
Machine Learning Engineer
November 2023 — October 2025
Genweb2 Limited
- Researching and Developing cutting-edge solutions to help people with communication disabilities.
- Working on a state of the art Vision Model to recognize sign language from video in real-time.
- Developing backend to serve Machine Learning Models.
- Automating a variety of tasks to reduce manual efforts.
- Designing and developing database and backend for Chat Communicator application.
- Optimized previous data processing methods.
- Developed methods to Augment video data to mitigate data constraint.
- Prepared Model development and deployment Documentation and Report.
Research Assistant
Spring 2023 — Present
Supervisor: Dr. Shafin Rahman.
Department of Electrical and Computer Engineering, North South University
Area of Research:
- Zero-shot Learning
- Computer Vision
Backend Developer
July 2022 — June 2023
Qavlo
- Contributed to develop backend and database for a financial service that facilitates international transaction to banks and mobile financial services.
Lab Instructor
Fall 2021 — Fall 2022
Department of Electrical and Computer Engineering North South University
Courses:
- CSE115L: Programming Language 1 Lab
- CSE311L: Database Systems Lab
Responsibilities include:
- Conducting lab classes for the undergraduate students and preparing lab materials.
- Implementing theoretical concepts with respective programming languages of the courses.
- Grading answer scripts and holding consultation hours for the students.

Education

North South University
Summer 2022 - Present
Dhaka, Bangladesh
- Masters of Science in Computer Science and Engineering
- Area of concentration: Intelligent System Engineering
North South University
Spring 2017 - Spring 2021
Dhaka, Bangladesh
- Bachelor of Science in Computer Science and Engineering
- Area of concentration: Artificial Intelligence
Moulavi Samsul Karim College
Batch 2016
Chhagalnaiya, Feni
- Higher Secondary School Certificate (HSC)
- Group: Science
Chhagalnaiya Academy High School
Batch 2014
Chhagalnaiya, Feni
- Secondary School Certificate (SSC)
- Group: Science

Skills

Programming Language

Python, C, MATLAB
Machine Learning

Object Detection, Image Classification, Action Recognition, Recommendation System, Retrieval-Augmented Generation, Data Visualization, OpenCV, Scikit-learn, TensorFlow, Keras, PyTorch
Backend & Database

FastAPI, Flask, Django, HTML, JavaScript, REST API, SocketIO, Redis, MySQL, PostgreSQL
Graphic Design

Adobe Illustrator, Adobe XD, Figma
Productivity Tool

Latex, Microsoft Office Suite, Google Workspace
OS & Version Control

Windows, Linux, Git, GitHub
Soft Skill

Teamwork, Leadership, Problem Solving, Adaptability
Language & Others

Language: Bengali (Native) and English
Other: Translation

Publications

Papers

Aggregated Fuzzy Signature Structures for Multi-class Medical Diagnosis
2024 International Conference on Neural Information Processing (ICONIP)
- Md Zakir Hossain, Zi Jin, Tom Gedeon, Amrijit Biswas, Ruchira Tabassum, Fahimul Hoque Shubho, Shafin Rahman
Abstract: The fuzzy signature structure has proven to be a dependable classifier by determining the compatibility and diversity of features of any complex structured data. One highly demanding arena for fuzzy structured applications is a medical diagnosis, such as COVID or SARS, where the aspects are difficult to categorize. A multitude of aggregation combinations can be adopted to obtain the highest accuracy of fuzzy signatures, which is what has been implemented in this paper. The number of classes was five, including patients with SARS, Normal, High blood pressure, Pneumonia, and COVID, where the comparison and significant differences between classes were analyzed. For several aggregation results, even the accuracy 100% was obtained, whereas, for a few combinations, the model did not work well, so it was necessary to select a suitable combination of aggregation to obtain a better outcome. Along with the implementation of the fuzzy signature model with mul- tiple aggregations, the K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) have also been executed for classification upon the identical fuzzy signatures, and validation of the proposed Fuzzy signature model has been established in terms of resultant accuracies. Approximately 98% precision has been observed for both the KNN test and the training score as the maximum, while the minimum score was 22% using combinations of aggregation. SVM also showed a maximum accuracy of 87%. So, statistical and computational validity checks proved the feasibility of the proposed fuzzy model.

Read more at:
ChatGPT-guided Semantics for Zero-shot Learning
2023 International Conference on Digital Image Computing: Techniques and Applications (DICTA)
- F. H. Shubho, T. F. Chowdhury, A. Cheraghian, M. Saberi, N. Mohammed and S. Rahman, "ChatGPT-guided Semantics for Zero-shot Learning," 2023 International Conference on Digital Image Computing: Techniques and Applications (DICTA), Port Macquarie, Australia, 2023, pp. 418-425, doi: 10.1109/DICTA60407.2023.00064.
Abstract: Zero-shot learning (ZSL) aims to classify objects that are not observed or seen during training. It relies on class semantic description to transfer knowledge from the seen classes to the unseen classes. Existing methods of obtaining class semantics include manual attributes or automatic word vectors from language models (like word2vec). We know attribute annotation is costly, whereas automatic word vectors are relatively noisy. To address this problem, we explore how ChatGPT, a large language model, can enhance class semantics for ZSL tasks. ChatGPT can be a helpful source to obtain text descriptions for each class containing related attributes and semantics. We use the word2vec model to get a word vector using the texts from ChatGPT. Then, we enrich word vectors by combining the word embeddings from class names and descriptions generated by ChatGPT. More specifically, we leverage ChatGPT to provide extra supervision for the class description, eventually benefiting ZSL models. We evaluate our approach on various 2D image (CUB and AwA) and 3D point cloud (ModelNet10, ModelNet40, and ScanObjectNN) datasets and show that it improves ZSL performance. Our work contributes to the ZSL literature by applying ChatGPT for class semantics enhancement and proposing a novel word vector fusion method.

Read more at:
Personality perception using scenario based stimulation and physiological signals
2023 IEEE International Conference on Systems, Man, and Cybernetics (SMC)
- A. Biswas, J. Li, F. H. Shubho, T. Gedeon, S. Rahman and M. Z. Hossain, "Personality Perception Using Scenario Based Stimulation and Physiological Signals," 2023 IEEE International Conference on Systems, Man, and Cybernetics (SMC), Honolulu, Oahu, HI, USA, 2023, pp. 1884-1891, doi: 10.1109/SMC53992.2023.10394366.
Abstract: Previous studies on automatic personality perception have primarily focused on a limited number of personality traits. However, in real-world situations, humans exhibit a wide range of personality traits. To overcome this limitation, a new methodology for automatic personality perception is proposed in this paper. This revised approach can predict various personality traits (17 traits) with satisfactory performance by utilizing physiological signals. The underlying concept is to stimulate participants with different emotional stimuli to elicit physiological responses in a specific scenario. Biomarkers such as Electroencephalogram (EEG), Skin Conductance, Blood Volume Pulse, and Pupil Dilation reflect an individual’s personality traits. Two experiments are conducted with different scenarios, including Image/Video Stimulation and Driving Simulation, to support this study. Based on the collection of data and validation of supervised learning models, Naive Bayes outperforms other classifiers explored in this research. EEG is the most effective signal for predicting personality, although combining other signals may produce similar results. Our method accurately predicts the 17 personality traits, demonstrating significant potential for clinical research.

Read more at:
Use of Social Media in Flood Assessment in Bangladesh
2022 IEEE 11th International Conference on Intelligent Systems (IS)
- F. H. Shubho et al., "Use of Social Media in Flood Assessment in Bangladesh," 2022 IEEE 11th International Conference on Intelligent Systems (IS), Warsaw, Poland, 2022, pp. 1-8, doi: 10.1109/IS57118.2022.10019640.
Abstract: Widespread floods are one of the most destructive natural phenomena that frequently occur in Bangladesh. It devastates human life, food security, shelter, and social and financial losses. Due to Bangladesh's current economic situation, installing and maintaining comprehensive flood gauges for national flood assessments will not be viable. Hence, online social media platforms can be a valuable avenue to get real-world data to perform flood assessments. This study investigates whether the abundance of data collected from social media in Bangladesh could be used to achieve a reliable and consistent flood assessment in Bangladesh. The data gathered from online platforms are related to flooding and exist in the form of videos, images, and text. The data collected is analyzed with a Machine Learning approach. The digital data of images and video frames were converted into numeric values using the VGG-16 architecture. A Convolutional Neural Network takes in the numeric data produced by the VGG-16 for classification. The classification made accuracy of 92% for the image-based model and 87% for the text-based model.

Read more at:
Real-time traffic monitoring and traffic offense detection using YOLOv4 and OpenCV DNN
TENCON 2021 - 2021 IEEE Region 10 Conference (TENCON)
- F. H. Shubho, F. Iftekhar, E. Hossain and S. Siddique, "Real-time traffic monitoring and traffic offense detection using YOLOv4 and OpenCV DNN," TENCON 2021 - 2021 IEEE Region 10 Conference (TENCON), 2021, pp. 46-51, DOI: 10.1109/TENCON54134.2021.9707406.
Abstract: This paper presents a computer vision-based system for traffic offense detection. The system detects traffic offenses such as speed limit violations, unauthorized vehicles, traffic signal violations, unauthorized parking, wrong-way driving, and motorbike riders without helmets. The traffic offense detection system consists of a pipeline of four different modules. These are a vehicle detection module, a vehicle classification module, a vehicle tracking module, and a traffic offense detection module. Vehicles on the roads are detected in the vehicle detection module using visual data such as live camera feed. Next, after the vehicles are detected, they are classified into different classes using a vehicle classification module. A vehicle tracking module is developed to track the vehicle as it moves through the traffic. Lastly, we have implemented a traffic offense detection module that analyzes traffic patterns and detects different types of traffic violations in real-time. The entire system is implemented using OpenCV Deep Neural Network (DNN) module. We have used YOLOv4 to detect vehicles on the roads with high accuracy. For motorbike riders without helmets, we have used a fast YOLOv4-tiny model. The DeepSORT algorithm is used to track vehicles in real-time. Obtained accuracies are 86% in YOLOv4 for vehicle detection and 92% in YOLOv4-tiny for helmet detection.

Read more at: