Gourab Bagchi
Hi, I am an AI Architect with over 9 years of combined professional experience across industry roles and advance postgraduate education in computer vision, natural language processing (LLMs), multimodal AI, and data science. Proficient in Python, R, and related tools, I specialize in predictive modeling and deriving actionable insights from complex datasets to drive informed decision-making.
View my LinkedIn Profile
Selected projects in data science, machine learning and NLP
RAG Research Chatbot
RAG based research article summarization chatbot.
LoRa fine-tune a SDXL model to generate a specific art styled images
Utilizing LoRa, fine-tune an SDXL model to generate art-styled images with precision and efficiency, enhancing artistic expression through targeted neural network adjustments.
Automated Recommendation System using LLM
LLAMA2-based Recommendation System for Process Safety Enhancement Utilizing LLAMA2, I developed a recommendation system for the petrochemical industry to enhance process safety. This system applies sophisticated algorithms and techniques to analyze industry-specific data, offering customized suggestions for improving safety measures and protocols.
Risk Prediction from Text Data
Developed a Risk Prediction model for a petrochemical industry by analyzing their past safety data. The model trained with the past text based HAZOP reports and vectorized with TF-IDF vectorization techniques, and for a given consequence statement the model can predict risk in a 5 point scale.
LLM Chatbot
In this project, I developed a chatbot using LLAMA2 and LoRA to leverage their respective strengths. LLAMA2 provides robust language modeling capabilities, allowing the chatbot to understand and generate natural language responses effectively. Additionally, LoRA enhances the chatbot’s conversational abilities by incorporating reinforcement learning techniques, enabling it to adapt and improve over time based on user interactions. Together, LLAMA2 and LoRA create a powerful and adaptable chatbot solution for a wide range of conversational tasks.
Multilevel Classification for Text Data
In this project I have developed a Novel methodology for classifying text data into specific levels/features.
Multivariate Data Analysis
Developing predictive models for workplace incidents and injuries, utilizing multivariate data analysis to assess various factors such as timing, location, frequency, severity, and risk. The aim is to extract insights from incident reports, near-miss observations, and workplace inspections to inform safety decisions and mitigate potential risks.
A Real Estate Chatbot using LLM and RAG
Developed a Real Estate Chatbot utilizing LLAMA2 for advanced language processing and integrating RAG for comprehensive knowledge retrieval, enhancing user interactions and property inquiries. This chatbot employs state-of-the-art AI technologies to provide personalized assistance and streamline the real estate experience.
Time Series Forecasting
Implemented time series forecasting using AR, MA, and ARIMA models to predict the number of injuries in a plant, aiding in proactive safety measures and risk management. Leveraging historical injury data, these models provide insights into potential future incidents, allowing for informed decision-making and prevention strategies.
