My academic journey has been a dynamic and enriching experience, spanning across different parts of the globe. From the vibrant streets of Pune, India to the culturally diverse campuses of The University of Glasgow, I've had the privilege to immerse myself in various educational environments, each offering unique perspectives and opportunities for growth. My pursuit of knowledge has taken me on a transformative path, shaping my understanding of the world and fostering a passion for continuous learning. Whether studying in bustling metropolises or serene academic settings, I've embraced every opportunity to expand my horizons and deepen my expertise in Computer Science and Mathematics. Through my educational endeavors, I've not only acquired valuable academic qualifications but also cultivated essential skills such as critical thinking, problem-solving, and collaboration. Each institution I've attended has contributed to my personal and professional development, equipping me with the tools and insights necessary to thrive in today's dynamic global landscape.

Django Blog App

  • Developed a comprehensive functional web application using Python and Django, incorporating user management, content creation, and deployment capabilities.
  • Implemented core functionalities such as user authentication, profile management, post-CRUD operations, pagination, and email integration for enhanced user experience.
  • Deployed and optimized the application on both Linux and Heroku platforms, integrating AWS S3 for image storage and securing the application with HTTPS.

Sportify | A Football Chatbot

  • Developed a Task-Oriented Dialogue System tailored for English Premier League Football.
  • Conceptualized and implemented a chatbot using Google Dialog Flow CX, effectively integrating real-time data by orchestrating Restful API calls through webhooks.
  • Deployed Python code in Google Cloud Functions to ensure consistent performance and uninterrupted.
  • Thoroughly assessed and refined the chatbot across diverse scenarios, leading to a successfully elevated accuracy from 80% to an impressive 96%.

Baxter Robot Playing Chess

  • Engineered a virtual simulation of a Baxter robot playing chess, demonstrating the ability to pick and place pieces in accordance with chess rules and the environmental dynamics of the board and table, mirroring human-like intelligence.
  • Concepts Used: ROS, VS Code, Gazebo, Jupyter Notebook, Python

Machine Learning on Colorectal Cancer Tissue Patches

  • Engineered a machine learning model in Python to cluster 5,000 colorectal cancer tissue patches, encompassing 9 distinct tissue types.
  • Utilized cutting-edge clustering algorithms, including K-means and Louvain.
  • Applied dimensionality reduction techniques, specifically PCA and UMAP.
  • Assessed model performance and accuracy using Silhouette and V Measure scores.

Identifying different types of Cell Nuclei in Cancer Samples using two CNNs

  • Crafted a sophisticated deep neural network architecture, achieving an impressive precision of 93% in identifying cancerous cells.
  • Leveraged Pytorch to construct an 8-layer deep convolutional neural network, implementing methodologies including class balancing, Stochastic Gradient Descent optimisation, regularization, and meticulous hyperparameter tuning via Ray-tune.
  • Utilized Captum library to unravel model insights and evaluations such as Grad-CAM, Saliency Map, and Layer Activation, facilitating comprehensive interpretation.

ROBOCUP Championship

  • Simulated a soccer tournament with NAO6 Robots in Webots using Python.
  • The Team consists of 3 Midfielders and 1 Goalkeeper.
  • Strategically defined team roles, including offensive, defensive, and goalkeeping tactics, harnessing distance parameters and varied movement patterns such as linear and curved trajectories.

Stress Detection using Facial Landmarking Techniques

  • Constructed a 17-layered Deep CNN model to determine the stress level of a human by examining facial expressions.
  • Generated training datasets using a high-resolution image/video.
  • Evaluated the model on diverse test data and achieved an accuracy of 80%.
  • Concepts Used: Haar-Cascades, Local Binary Pattern Histogram, CNN & SVM.