About Me

I work as a reseach engineer at Fujitsu Research of America (FRA). I earned a Ph.D. from the Department of Computer Science and Engineering, The Ohio State University (OSU) supervised by Prof. Rajiv Ramnath. My research focuses on the application of in the application of machine learning, especially in Natural Language Processing (NLP). Before joining OSU, I obtained my B.Sc. in CSE from Bangladesh Universiy of Engineering and Technology (BUET). I also worked at Kona Software Lab Ltd at the Dhaka site for 2 years.

I have experience in employing neural information retrieval (IR) and reinforcement learning (RL) techniques to develop retrieval-guided response generation (NLG) models for conversational agents. At FRA, I have designed and developed a multi-agent system (RAG, Coding) for an AI chatbot to analyze tabular and textual data [Python, LangChain, LangGraph, Ollama, Streamlit, Chroma].

Publications

• Efficient and Interpretable Information Retrieval for Product Question Answering with Heterogeneous Data, Biplob Biswas, Rajiv Ramnath. In Proceedings of the Seventh Workshop on e-Commerce and NLP @ LREC-COLING, May 2024.

• Improving Response Automation for Customer Communication through Document Categorization and Information Retrieval, Biplob Biswas. Doctoral dissertation, The Ohio State University, 2023.

• Retrieval Based Response Letter Generation For a Customer Care Setting, Biplob Biswas, Renhao Cui, Rajiv Ramnath. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics (NAACL): Human Language Technologies: Industry Track, July 2022.

• TransICD: Transformer Based Code-wise Attention Model for Explainable ICD Coding, Biplob Biswas, Hoang Pham, Ping Zhang. In International Conference on Artificial Intelligence in Medicine (AIME), Jun 2021.

• Towards Simulating Non-lane Based Heterogeneous Road Traffic of Less Developed Countries, Q. M. Alam, B. Sarker, Biplob Biswas, K. H. Zubaer, T. Toha, N. Reza and ABM A. Al Islam. In ICT4S, May 2018.

Projects

• Designed and developed a multi-agent system (RAG, Coding) for an AI chatbot to analyze tabular and textual data. [Python, LangChain, LangGraph, Ollama, Streamlit, Chroma]

• Designed a sequence generation model for predictive process monitoring.(Python, PyTorch)

• Developed an ML-driven web tool for process data analysis and visualization.(Python, Streamlit)

• Designed and developed a knowledge-grounded response generation framework to automatically answer user queries in a customer care setting (Real-world data) and deployed it as an API service. [AWS, Docker, Python, PyTorch, HuggingFace]
  • Employed neural information retrieval (IR) techniques to collect knowledge from historical unstructured data and blended that information into generated responses through a fine-tuned GPT-2 model.
  • Formulated a novel ranking mechanism for better response selection from multiple generations.
  • The model generates 12% more informative and 15.3% more semantically similar responses than that of vanilla GPT-2 (contemporary state-of-the-art).

• Developed a paraphrasing model and corresponding web service to introduce diversity in the business response template. [AWS, Docker, Python, Python]

• Designed and developed a deep-learning model - TransICD (Code) to predict ICD codes from a patient discharge summary. [Python, PyTorch]
  • The model achieves stable micro-AUC and micro-F1 scores of 0.92 and 0.64 respectively with 1.1% and 2.5% increases over the corresponding metrics of LEAM (contemporary state-of-the-art).

• Developed a tool to extract and analyze pre-installed applications in Android Firmware. [Python, MySQL]

• Developed the front-end of NexusPay a leading Android payment application (1M+ Downloads) of Bangladesh. [Android, Java, SQLite]
  • Utilized REST APIs for backend communication and implemented various transaction interfaces including NFC, QR-code, and online payment.
• Developed a Nexgo POS application to facilitate the transaction from an Android payment application. [C]