Kajol

Natural Language Understanding (NLU), Natural Language Processing (NLP), Named Entity Recognition (NER), Sentiment Analysis, Topic Modeling, Text Classification, ChatBot, Dialogue Manager, Streamlit | Feb. 2023 - May 2023 [code]"

● Developed an advanced chatbot system with natural, fluid conversations, leveraging NLU techniques and incorporating a the re-ranking mechanism for human-like content retrieval. ● Integrated feedback analysis to adapt conversations based on user mood, ensuring personalized and seamless transitions.

PostgreSQL, Streamlit | Feb. 2022 - May 2022 [code]"

•Developed a movie information system with SQL database, E/R diagram, relational schema, data loading, BCNF normalization, and optimized SQL queries using triggers.
● Developed a ReactJS-powered website specializing in query visualization.

React, Django Face, Jupyter Notebook, PowerBI | Aug. 2022 - Dec. 2022 [code]"

• Developed an application to deduce the severity of accidents.
• Concepts: Data Cleaning/Processing, Exploratory Data Analysis, Applied 5 data models Logistic Regression, K Nearest neighbors, Decision Tree, Random Forest Classifier, Artificial Neural Network
• Created the react app and used Microsoft Power BI for visualization and to display the analysis on the dataset.

Numpy, Scipy, Sklearn, Matplotlib, Pickle, Torch, Jupyter Notebook | Aug. 2022 - Dec. 2022 [code]"

I worked on various Mini Projects exploring/experimenting various ML algorithms. Please find below the details for individual projects:
Project_01 :
Problem: The project is about different classification and regression techniques. There are six sub-problems under this project.
•Experiment with Gaussian Discriminators by implementing Linear Discriminant Analysis (LDA) and Quadratic Discriminant Analysis (QDA)
•Linear Regression implementation
•Experiment with Ridge Regression
•Using Gradient Descent for Ridge Regression
•Non-linear Regression
•Compare and interpret the results of different classification and regression methods

Project_02 :
Problem: Handwritten Digits Classification using Multilayer Perceptron (MLP) Neural Network
Approach:
•Implemented MLP Neural Network from scratch to classify handwritten digits from MNIST dataset (achieved test accuracy of 93.45%)
•Used Feed Forward and Back Propagation to implement Neural Network
•Experimented the effect of regularization in the bias-variance trade-off
•Used the same neural network for more challenging face dataset and compared it with Deep Neural Network and Convolutional Neural Network using the Tensorflow library

Project_03 :
Problem: The project is about Support Vector Machine (SVM) and mulit-class logistic regression.
Approach:
•Implemented Logistic Regression (using one-vs-all strategy) to classify handwritten digit images into correct corresponding labels. In particular, built 10 binary-classifiers (one for each class) to distinguish a given class from all other classes.
•Developed multiclass logistic regression classifier and SVM for handwritten digits classification on the MNIST Dataset

C++ | Aug. 2022 - Dec. 2022 [code]"

● Applied TCP/IP protocol and socket programming. Developed a client-server chat app enabling multiple users to log in, establish identity, and communicate via a central server. Utilized socket programming to establish robust connections, ensuring reliable message transmission.
● Implemented a buffering system for message storage and retrieval, especially for offline clients.Conducted thorough testing in a live production environment to ensure accuracy and efficiency under real-world conditions.

Python 3, Tensorflow, Keras, Matplotlib, OpenCV | July 2018 - Dec. 2018 [report]

• Built a tool to detect autism at a pre-clinical stage so that proper intervention can be provided for development of autistic children.
• Did a comparative analysis of three machine learning approaches of autism detection using eye tracking, which are, CNN, DNN, and non-neural network methods-SVM and Random Forest Classifier.
• An accuracy of 73.54% was obtained using CNN, 78.86% using DNN, and 63.07% and 65.22% for SVM and Random Forest Classifier respectively.

Python 2.7, NumPy, OpenCV3, Plotly-express, Scikit-learn, Freenect, Scipy, SimpleCV | Jan. 2018 - May 2018 [report]

• Developed an application for recognizing hand gesture for controlling the VLC media player.
• Concepts: Data extraction(using HoG feature), classification(Multiclass-SVM).
• The method which employs depth thresholding and SVM to recognise gestures, achieved an average accuracy of 95.83%.