Skip to content

PROJECTS UNDERTAKEN


Water Solubility Prediction

Project Summary

Developed an AI-driven solubility prediction model with water that reduced prediction time from 1–2 days to just seconds, enabling chemists to accelerate compound screening and decision-making processes by over 95%.

Techstacks Used

NumPy, Pandas, XGBoost, Chembert, GNN, MPNN, Stremalit,

Water Solubility Prediction

Multi-Solvent Solubility Prediction

Project Summary

Designed and developed an advanced web framework to accurately predict compound solubility across 25 diverse solvents and temperatures. This framework integrates compound physical properties with the ChemBERT model and is further enhanced by a Message Passing Neural Network (MPNN) to capture complex molecular interactions. Leveraged generative AI techniques for improved predictive accuracy, addressing challenges in solubility estimation for drug discovery and materials science. The solution was deployed using Streamlit, enabling intuitive visualization, and and also check applicability domain with t-SNE analysis

Techstacks Used

Scikit-learn, NumPy, Pandas,

Data-Pipeline-Project

Project Summary

The project involves constructing a data pipeline with Python scripts and , encompassing data fetching from Pubchem ( A big database for chemical compounds) processing and stored for further analysis and comparison with other predictive model.

Techstacks Used

Python ,Beutifulsoup, Pubchempy , Tableau

Talk to Research Paper using RAG

Project Summary

Retrieval-Augmented Generation (RAG) model-based chatbot application. The chatbot uses Ollama to help students answer queries by orchestrating a flow through various Research papers of their intrest and displaying the results.

Techstacks Used

Deepseek, RAG , Python

Cancer image classification

Project Summary

Developed a deep learning-based system to predict the presence of cancer in images across four types (brain, breast, skin, lungs) using a convolutional autoencoder with TensorFlow and Keras. The project was presented through a Streamlit web framework, offering an interactive platform for demonstrating the model's capabilities in cancer image classification

Techstacks Used

Convolutional Autoencoder,Anomaly detection,Image analysis, Python, streamlit

Twitter Data Exploration and Visualization

Project Summary

Cleaned and visualized a complex crime dataset using Tableau Prep and Tableau Desktop facilitating better decision making through reliable insights.

Techstacks Used

Tableau Prep, Trifacta Data Prep, Kafka, Data Profiling, Data Cleaning and Uncleaning processes.