Research Experience

Research Assistant
Supervisor: Dr. Laith H. Baniata
Assistant Professor, Gachon University, South Korea
Email: laith@gachon.ac.kr
June 2024 - Present

• Conducted research on "Investigating the Predominance of Large Language Models in Low-Resource Bangla Language Over Transformer Models for Hate Speech Detection: A Comparative Analysis." This work was 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.
• Carried out research on "SentimentFormer: A Transformer-Based Multi-Modal Fusion Framework for Enhanced Sentiment Analysis of Memes in the Under-Resourced Bangla Language." This work was 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

Application Developer (AI/ML)
Dexian Bangladesh LTD.
May 2024 - Present

• Developed the RFPMatcher, an advanced RAG solution to extract key information and summaries from Request for Proposal (RFP) documents using domain-specific multitask prompts. The system evaluates proposal responses and incorporates a Past Experience Matcher Score to compare bids, aiding in the prediction of potential win or loss outcomes. Additionally, the system generates a Table of Content, which serves as a structured guide for writing proposals for new bids.
  Technologies used: Python, LlamaIndex, Azure OpenAI, AlloyDB, CouchDB, SQL, React JS, Tailwind CSS, FastAPI
• Designed the CaseAligner, an LLM-powered application that generates tailored PowerPoint presentations for case studies based on specific Practice and Industry categories using internal project data. The application features a dynamic chat interface for interacting with individual slides, advanced summarization functionality for concise summaries of entire presentations or specific slides, and a robust search interface to locate information across generated case studies. Additionally, it enables users to regenerate specific slides for customization, automating content creation.
  Technologies used: Python, LlamaIndex, Azure OpenAI, AlloyDB, React JS, FastAPI
• Worked on KnowledgeEngine, an LLM-based Multi-Document innovative RAG Q&A system for retrieving early retirement insurance information. It analyzes internal reports and legal guidelines to provide context-aware answers with references. The system uses task-specific prompts to calculate retirement insurance costs, benefits, and tax implications. Each user has a dedicated knowledge base to ensure data security and privacy for uploaded information.
  Technologies used: Python, LlamaIndex, Azure OpenAI, AlloyDB, React JS, Tailwind CSS, FastAPI
• Implemented Org Info, a vision-language model (VLM)-based application designed to simplify the extraction, management, and querying of organizational hierarchy data. The application automates the extraction of hierarchical information from uploaded images of organizational charts, stores the data in a structured database, and provides a chat interface for intuitive user interaction.
  Technologies used: Python, LangChain, Azure OpenAI, OpenCV, SQL, SQL Alchemy, React JS, FastAPI
• Designed AgentDexi, an LLM-based multi-agent system and RAG solution that analyzes job demand across various companies to provide actionable insights. This solution empowers technical recruiters to optimize their hiring strategies by aligning recruitment efforts with current industry trends and enables customization to meet specific needs for more targeted and effective outcomes.
  Technologies used: Python, LangChain, CrewAI, Azure OpenAI, React JS, Tailwind CSS, FastAPI
• Built Dexian Innovation Insights, an LLM-based multi-agent recommendation system that identifies emerging trends within a company's internal project data, compares them with external data, and generates detailed reports summarizing the insights with interactive graphs and charts. This system helps the company stay ahead of technological trends and supports decision-makers by offering data-driven guidance on project directions and potential areas for innovation.
  Technologies used: Python, LangChain, CrewAI, Azure OpenAI, React JS, Tailwind CSS, FastAPI