I'm Kanan Pandit
Master's Student in Data Science
I am currently pursuing a Master’s degree in Data Science, with a strong focus on machine learning, data analytics, and AI technologies. With a solid foundation in statistical modeling and data manipulation, I specialize in transforming complex data into actionable insights. I am passionate about solving real-world problems through innovative, data-driven approaches and committed to continuous learning.
Ramakrishna Mission Vivekanada Educational and Research Institute, Belur
2024-2026 | Pursuing
WBUTTEPA,Kolkata
2020-2022 | Completed
Vidyasagar University,Medinipur
2017-2020 | Completed
Golar Sushila Vidyapith High School,Golar,Medinipur
2015-2017 | Completed
Golar Sushila Vidyapith High School,Golar,Medinipur
2015 | Completed
This project focuses on preprocessing the EMNIST dataset and applying various classification models—Logistic Regression, Softmax Regression, KNN, Decision Tree, Random Forest, and SVM—for handwritten character recognition. We compare model performance to identify the most effective approach and address challenges related to handwriting variation, offering recommendations for improved recognition systems..
View Code
This project involves applying and comparing various regression techniques—including linear, polynomial, gradient descent methods, and regularization (Ridge, Lasso, ElasticNet)—to predict sales based on advertising data. The goal is to evaluate model performance using metrics like MAE, MSE, R², and computational time, with Polynomial Regression (Degree 3) identified as the most accurate and efficient model.
View Code
A dynamic and interactive portfolio designed to highlight my journey in data science. Showcases core projects involving machine learning, data visualization, and statistical analysis, with an emphasis on real-world problem-solving and continuous learning.
View Code
This project demonstrates the setup of a distributed H2O cluster across two machines and the execution of a machine learning task, showcasing parallel processing and scalable model training in a multi-node environment.
View Code
This project involves configuring a multi-node Apache Spark cluster and executing distributed machine learning tasks, highlighting Spark's capability for large-scale data processing and model training.
View Code
Implemented distributed machine learning using H2O AutoML to predict wildfires, leveraging multiple nodes for scalable and efficient model training.
View Code
Developed a Smart Control Hub using Mediapipe for real-time hand tracking to enable gesture-controlled volume, brightness, virtual mouse, and presentation control with visual feedback for intuitive interaction.
View Code
Implemented a CycleGAN model to transform real-world photos into Studio Ghibli-style images, preserving key features like facial structure, and evaluated the results qualitatively during training.
View Code
Conducted a comparative analysis of various image filtering techniques and developed a hybrid image generation algorithm to enhance image processing outcomes.
View Code
Implemented Harris corner detection to identify key points in images and performed feature matching to align and compare images effectively.
View Code
Developed image stitching for panorama creation, image alignment like CamScanner, and depth estimation from stereo images using OpenCV and a custom Harris corner detector without relying on OpenCV’s built-in functions.
View Code
This project fine-tunes BERT for binary sentiment classification on the IMDB dataset and investigates its vulnerability to adversarial attacks. Developed multiple attack methods using semantically valid word substitutions to fool the model, evaluated attack success, and improved robustness through adversarial training with filtered adversarial examples.
View Code
Python, C, R
MySQL, Hadoop, Spark, Power BI
NumPy, Pandas, Scikit-learn, PyTorch
