KUNAL AGARWAL

A Data Science aspirant and a self-taught Machine Learning practitioner eager to derive inferences from the data, create informative visualizations to help in a clearer understanding of the data and add predictive capabilities to the model built.

An experienced Software Development Analyst in Java and SQL of 3 years and 9 months with the curiosity and knack to take on new and challenging roles and responsibilities.

Built a Real Time Face Mask Detector using my own Webcam images using OpenCV and CNN

• Trained a CNN on 3 sets of images with mask wore properly, mask below nose and without a mask using Webcam.
• Used OpenCV and Mediapipe to detect the face and crop the face bounding box for data collection.
• Used OpenCV and Mediapipe to detect if face is covered by hand displaying the result.
• Obtained 90% Categorical Validation Accuracy on my own face and 78% accuracy on external data.

Built a Cat vs Dog classifier hosted on Heroku via Streamlit and Github on Dogs-vs-Cats dataset on Kaggle

• Analyzed Dogs-vs-Cats dataset from Kaggle and performed data preprocessing on Images.
• Created a CNN model and trained it on 8000 images to achieve 89% accuracy on test set.
• Deployed the model on Streamlit and hosted it on Heroku via Github. Can be tested by uploading an image of a cat/dog.
• Model can be found on: http://cat-or-dog-streamlit.herokuapp.com/

Predicting Survival on RMS Titanic using Random Forest and Hyperparameter Optimization

• Analyzed RMS Titanic data from Kaggle dataset using statistical and visualization tools in Python.
• Trained various models like Logistic Regression, Decision Trees, SVM, KNN and Random Forest on the data.
• Used GridSearch CV to optimize hyperparameters of Random Forest classifier to attain 80% accuracy on the Kaggle leaderboard.

Handwritten Digit Recognition on MNIST Dataset using Pipeline

• Analyzed the MNIST Dataset for classifying handwritten digits to the correct labels.
• Trained Logistic Regression model on the data and achieved a baseline accuracy of 84%.
• Created a Pipeline to cluster the data using KMeans and modified the input images to the distance of these images from the clusters and gave that as input to the Logistic regression model improving the accuracy to 90.8%.

Creating an SMS Spam Filter from the UCI SMS Spam Collection Dataset using Natural Language Processing

• Analyzed SMS Spam Collection Dataset on University of California Irvine Machine Learning Repository using various plotting tools in Python.
• Tokenized the data removing the punctuation and stopwords.
• Created a bag-of-words model and computed the TF-IDF weights of the words for the model.
• Created a Pipeline to calculate the TF-IDF weights and train the model to attain an accuracy of 96%.

• Python for Data Science and Machine Learning Bootcamp, Udemy,

Father’s Name: Mr. Dinesh Kumar Agarwal

Birthday: 21/11/1994

Gender: Male

Marital Status: Single

Nationality: Indian

Permanent Address: H.No.151A, Friends Enclave, Gali No.10, Vaishali Colony, Kashipur, Uttarakhand-244713 (Distt: Udham Singh Nagar)

Mobile Numbers: +91-8266047129 / 7055243493

E-mail: agarwalkunal12@gmail.com

I, Kunal Agarwal, hereby declare that the information contained herein is true and correct to the best of my knowledge and belief.

Kunal Agarwal