Research Experience
Remote Research Assistant
Supervisor: Dr. Laith H. Baniata
Research Professor, Gachon University, South Korea
Email: laith@gachon.ac.kr
June 2024 - Present
1. Towards Robust Chain-of-Thought Prompting with Self-Consistency for Remote Sensing VQA: An Empirical Study Across Large Multimodal Models
- ■ The research was initiated with support from the National Research Foundation of Korea (Grant No. NRF-2022R1A2C1005316), funded by the Ministry of Science and ICT.
2. Analyzing Diagnostic Reasoning of Vision-Language Models via Zero-Shot Chain-of-Thought Prompting in Medical Visual Question Answering
- ■Investigated research supported by the National Institute of Health (NIH) project in South Korea (Project No. 2024ER080300), funded by the Basic Science Research Program through the National Research Foundation of Korea (NRF) under the grant NRF-2022R1A2C1005316.
3. Investigating the Predominance of Large Language Models in Low-Resource Bangla Language Over Transformer Models for Hate Speech Detection: A Comparative Analysis
- ■ Conducted research supported by the Basic Science Research Program of the National Research Foundation of Korea (NRF), funded by the Ministry of Science and ICT under the grant NRF-2022R1A2C1005316.
4. SentimentFormer: A Transformer-Based Multi-Modal Fusion Framework for Enhanced Sentiment Analysis of Memes in the Under-Resourced Bangla Language
- ■ Carried out research supported by the Basic Science Research Program of the National Research Foundation of Korea (NRF), funded by the Ministry of Science and ICT under the grant NRF-2022R1A2C1005316.
Professional Experience
Senior Application Developer
Dexian (Bangladesh) Limited.
July 2025 - Present
Conversational Agent Platform for Financial Document Assistance: ShareFlow Agent
- ■ Developed and deployed a Microsoft SharePoint-integrated ReAct Agentic RAG system that enables users to create personalized agents for retrieving information from recruiting processes and internal financial document repositories.
- ■ Built a Multimodal OCR Agent with custom tools to process and extract information from a wide variety of unstructured, scanned documents, with the capability to autonomously select the appropriate tool and use a fallback mechanism.
- ■ Implemented session-based chat functionality with long-term memory management, ensuring each user's conversations with individual agents are kept separate, with full history retention for context-aware interactions.
- ■ Generated leading questions based on the agent's instructions and description to guide user interactions.
- ■ Integrated a user interface to display the list of agents created by the user or shared with them, along with an update feature that allows users to modify existing agents.
- ■ Designed a sharing functionality that allows users to share their agents either publicly with all app users or privately with a specific group of people for collaborative use.
- ■ Guided and mentored junior application developers to support their technical development, promote best practices, and ensure the delivery of high-quality, maintainable solutions.
- ■ Achieved approximately 63% cost reduction by streamlining custom agent usage, enabling 80 Sales Managers to handle 50+ daily interactions while replacing the existing SharePoint Agent.
- ■ Optimized RAG search accuracy by 96%, automating document content extraction and minimizing human effort.
- ■ Tech Stack Used: Microsoft SharePoint, Python, LlamaIndex, Azure OpenAI, Azure SQL, Azure Functions, AlloyDB, React JS, FastAPI
Application Developer
Dexian (Bangladesh) Limited.
May 2024 - July 2025
Organizational Intelligence Role Placement System: Org Info
- ■ Implemented a multimodal agent for extracting organizational hierarchical information from organograms, utilizing in-context learning with tree-of-thought prompting to preserve the correct parent-child structure.
- ■ Applied multipath reasoning to resolve conflicts and ambiguities in relationships for accurate role placement and stored the extracted hierarchical information in a relational database.
- ■ Designed an LLM-based agent that converted natural language queries into SQL using few-shot learning and self-consistency with chain-of-thought prompting, which enabled contextual reasoning to accurately retrieve relevant organizational data and integrated the results into the OrgChart framework for hierarchical visualization.
- ■ Developed a dynamic Agentic RAG-guided chat interface that enabled users to interact with a specific organizational hierarchy by utilizing predefined query types, roles, and goals, and delivered context-aware, natural language responses.
- ■ Set up scheduled jobs to fetch organizational data from Bullhorn every 30 days and visualized organizational information in OrgChart.
- ■ Streamlined organizational hierarchy retrieval for Account Managers by reducing search time by 92%, enabling faster access to relevant data without querying the entire Bullhorn database.
- ■ Tech Stack Used: Python, LangChain, LangGraph, Azure OpenAI, OpenCV, Azure SQL, React JS, FastAPI
Next-Gen Proposal Automation Engine: RFPMatcher
- ■ Developed a Retrieval-Augmented Generation (RAG) solution using Chain of Thought prompting to extract key information from Request for Proposal (RFP) documents.
- ■ Built a Past Experience Matcher system that uses Automatic Chain-of-Thought prompting alongside in-context learning and preset questions to extract requirements from RFPs, then matches them against a master database of prior proposal responses.
- ■ Enabled the system to generate Yes/No decisions with detailed explanations of how similar requirements were addressed in the past, aiding in the prediction of potential win/loss outcomes for new proposals.
- ■ Generated dynamic Tables of Contents (TOC) based on extracted key information and historical experience to streamline and structure the proposal writing process for new bids.
- ■ Reduced manual review time by 75%, increased proposal accuracy, and improved decision-making speed through automated extraction and predictive insights, empowering Proposal Managers to focus on strategic bid development.
- ■ Tech Stack Used: Python, LlamaIndex, Azure OpenAI, AlloyDB, CouchDB, React JS, FastAPI
Automated Presentation Insights Generator: CaseAligner
- ■ Built an LLM-based application using zero-shot prompting to generate PowerPoint presentations for case studies based on selected practice areas and industries.
- ■ Implemented an interactive chat interface allowing users to query specific slide content and receive instant contextual responses.
- ■ Developed comprehensive search functionality to locate information across all generated case studies.
- ■ Created export capabilities for downloading slides in company's official template.
- ■ Designed an admin panel for authorized users to download and edit the knowledge base of case studies.
- ■ Significantly accelerated demo preparation by reducing slide development time by nearly 90%, enabling Sales Representatives to prioritize client interactions and drive deal closures.
- ■ Tech Stack Used: Python, LlamaIndex, Azure OpenAI, React JS, FastAPI
Legal Document Information Retrieval System: KnowledgeEngine
- ■ Developed an LLM-based, multi-document RAG Q&A system for internal document information retrieval.
- ■ Implemented a chat conversation interface with document page references for information sources.
- ■ Maintained session-based dedicated knowledge bases to ensure data isolation and user-specific context management.
- ■ Created an admin panel with document upload functionality and comprehensive document management capabilities.
- ■ Tech Stack Used: Python, LlamaIndex, Azure OpenAI, AlloyDB, React JS, FastAPI
Smart Recruitment Analytics Tool: AgentDexi
- ■ Designed an LLM-based multi-agent system that scrapes job postings and analyzes demand trends to generate customer intelligence reports with graphical charts.
- ■ Developed an RAG solution to identify technological trends in job descriptions across external companies.
- ■ Created interactive graphical charts and dashboards that minimized analysis time by 80%, facilitating Technical Recruiters to quickly interpret data insights and make timely, data-driven hiring decisions.
- ■ Tech Stack Used: Python, LangChain, CrewAI, Azure OpenAI, React JS, FastAPI