Suvojeet Dutta
Senior Data Scientist (Gen AI) @ IBM || Multi-Agent System & RAG || LLM Fine-Tuning (Full FT/LoRA/QLoRA/DoRA) || Ex-Persistent || Ex-ISI (R&D) || Ex-TCS || M.Tech in CSE
- Role
- Senior Data Scientist at IBM
- Location
- Kolkata, WB, IN
- LinkedIn followers
- 500 followers
About Suvojeet Dutta
Currently working as a Senior Data Scientist with ~7 years of experience at IBM and developing Agent Builder platform.I have developed LLM driven applications including multimodal RAG chatbot and have fine-tuned multiple LLMs/SLMs as a Gen AI Data Scientist at Persistent Systems. I have worked as an MLE (Project Linked) at Indian Statistical Institute (Kolkata) and Data Scientist at Tata Consultancy Services on latest technologies in the field of Data Science/ Machine Learning/ Deep Learning/ Computer Vision/ NLP.Skillset - LLM Fine-Tuning (LoRA, QLoRA), SLM Fine-Tuning, Continual Pretraining LLMs Text based RAG, Multimodal RAG ChatbotAgentic Framework - Langchain, LangGraph, AutoGenProgramming Language: Python, C++(Basic) Data Visualization: Matplotlib, Seaborn, PlotlyNumerical Computation: NumPy, SciPyData Analysis: PandasML Libraries: Scikit-Learn, XGBoostDL Libraries: Keras, TensorFlow, PyTorchDatabases: Mysql, PostgresSQL, MongoDB, Cassandra Cloud Platforms: GCP, AWSVersion Control: GitOthers: REST, FastAPI
Experience
Senior Data Scientist
Jul 2025 — Present · Kolkata, IN
Developing Agent Builder to create Single/Multi-Agent Systems using Snowflake Cortex, LangGraph with MCP and Long-Term Memory Management Designed and implemented a modular Multi-Agent architecture leveraging LangGraph to coordinate complex task flows across specialized agents. Integrated multiple task-specific tools as Model Context Protocol (MCP) servers, utilizing SSE, Stdio, and Streamable HTTP interfaces for diverse operational requirements. Established short-term memory management via PostgresSaver, enabling session-specific state tracking through a persistent state graph. Architected long-term memory capabilities using PostgresStore, supporting structured memory types including Semantic, Episodic, and Procedural for enhanced agent context and learning. Developed a user-friendly interface using Streamlit, enabling real-time interaction and visualization of multi-agent workflows and memory states.
Education
University of Calcutta
Bachelor of Science - BSc, Computer Science
2012 — 2015
University of Calcutta
Master of Technology - MTech, Computer Science
2017 — 2019
University of Calcutta
Master of Science - MSc, Computer Science
2015 — 2017
Find verified contacts for anyone on LinkedIn
Unifers gives sales teams verified emails and direct dials, enriched profiles, and outreach that lands in the inbox.
Free plan included · No credit card required
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.