Research

Research Experience

Ongoing

• Astha: An AI-Based Mental Health Companion for Youth [Astha]

In Collaboration with the undergraduate students of BUET 

• • The objective is to develop a holistic platform for the mental health support of young generation

  • Selected for Startup incubation support under ICSETEP project, University Grants Commission of Bangladesh

  • We are planning for features like Psychologist sessions via personalized & customized video call and other options, AI chat & monitoring (conversational text-based and voice assistant), Addiction control games, and Community forum

  • Mental health challenges faced by Bangladeshi students: Cause & Effect. [Report]

• Development of an Accessible Wheelchair with Intelligent Caregiving Features [C.A.R.E Demo]

In Collaboration with college students and the undergraduate students of BUET

• • We aim to understand the needs of individuals with paraplegia, tetraplegia, diplegia, hemiplegia, and monoplegia, identify customized design alternatives, develop prototypes, and perform user-centered evaluation

  • Currently, we have features like head/hand-gesture controlled mobility, intelligent safety & emergency stop, real-time IoT-enabled health & environment monitoring, etc.

• Conversion of noisy injured face image to real face image 

     In Collaboration with the undergraduate students of BUET 

• • The goal of this research work is to convert fractured or corrupted face image (this can be happened due to different kinds of accidental and criminal issues such as fire burn, road accident, decomposition of dead body, etc.) to actual face image.

  • In this work, we are using synthetic injured images and corresponding real face images scraped from from public websites

  • We aim to train U-Net, Resnet, GANs, and other models and do comparison among in terms of performance among those models

Past

• Evaluating the Impact of Information Source in Flagging Content [Technical Paper]

Under the supervision of Dr. Hana Habib, Assistant Teaching Professor, Carnegie Mellon University at Center for Applied Research on Targeted Violence (CARVt)

• • Social media platforms have implemented significant efforts to flag dangerous or unverified content by labeling those content.

  • Those labeling can impact users' perceptions and interactions with the labelled content.

  • The objective of this research study is to explore which information source labels are most effective in warning users about dangerous or unverified content that they may encounter.

• Social media data analysis for understanding the condition of mental health and well-being of  Bangladeshi adolescent students [Thesis Book]

M.Sc. Thesis, Under the supervision of Dr. Mohammed Eunus Ali, Professor, Department of CSE, BUET

• • Analyzing self-reported survey data for understanding students’ mental health disorders

  • Identifying topics in Bangla by clustering transformer-based word embeddings [paper]

  • Exploring insights related to mental health disorders (i.e., anxiety, stress, depression, and suicidal ideation) from social media data

• An AI based medical chatbot for adolescent health support in Bangla language

B.Sc. thesis, Under the supervision of Dr. Mohammed Eunus Ali, Professor, Department of CSE, BUET

• • The main objective of this work was to develop a chatbot which would act like a personal assistant to mitigate the problems faced by adolescents during adolescence phase. Again open discussion related to adolescence sexual and reproductive heath is considered as taboo in Bangladesh which was another strong motivation of our research work.

  • We performed feasibility analysis to understand the opportunities of chatbot in this specific domain. [The paper is accepted for publication in CHI '21. (paper)]

  • We deployed our chatbot in Facebook messenger where we used NLP tool wit.ai and flask as the back-end.

  • We commercialized the chatbot and renamed "Susatho AI". [paper] 

• Understanding Data Privacy Concerns in the Shared Use of LLM

     In Collaboration with the undergraduate students of BUET 

• • RQ1: What are users’ perceptions and attitudes toward using a shared ChatGPT+ account?

  • RQ2: How do users behave while using a shared ChatGPT+ account?

  • RQ3: In what ways does sharing a ChatGPT+ account impact users’ data privacy?

  • We designed, developed, and commercialized a tool named Ziho AI addressing all challenges in shared use of LLMs

• Exploring opportunities for remote learning platforms for individuals with ASD to facilitate special needs education in Global South (e.g., Bangladesh) during the COVID-19 pandemic [Preprint]

Under the supervision of Dr. Mohammed Eunus Ali, Professor, Department of CSE, BUET

• • The main objective is to identify challenges and problems faced by individuals with ASD, their teachers and parents in online education through zoom platform and to gauge the efficacy of such pedagogical methods compared to conventional special needs education.

• A Novel Approach of Generating Robust Session Identifier for Preventing Session Hijacking Attacks [Preprint]

Under the supervision of Dr. Md. Shohrab Hossain, Professor, Department of CSE, BUET

• • Our main objective is to implement a preventive mechanism in application layer which ensures session integrity. 

• Sigma Promoter Identification & Classification in E. coli Bacteria from Genome Data using Sequence Model based Approach

Under the supervision of Dr. Atif Hasan Rahman, Assistant Professor, Department of CSE, BUET

• • The objective of this research is to implement a computational tool that will identify and classify promoters which is responsible for transcription in E. coli bacteria automatically from DNA sequence.

  • We have formulated a novel sequence model based classification method using Keras. We have implemented simple RNN, LSTM, & Bi-LSTM based approaches. So far, our study is giving comparable results with the state-of-the-art works. We are currently doing rigorous testing on this model.