Pursuing Masters' in Computer Science at North Carolina State University.
I'm versatile and inquisitive mind engineer with a passion for application development and Machine Learning problem solving. I believe in building technology that empowers mankind leading to a smarter and better future.
I'm proficient with Python and worked on multiple machine learning projects to enhance my understanding of the field. I am familiar with multiple machine learning, deep learning and statistical analysis libraries like scikit-sklearn, keras, tensorflow, OpenCV, pandas, and numpy.
Looking for Full-Time opportunites in Software Development, Machine Learning, or Data Analytics.
Apart from Technology, my passion dwells in Photography, Writing and Traveling. I love to interact with people from various cultures, of different countries and sharing experiences as I believe there's something to learn from every experience.
You can connect with me through my social media accounts shown below.
• Used kaggle dataset of over 150,000 comments to classify them in their toxicity nature i.e. if they have abusive language, insulting or represent negative sentiment.
• Built 5 models, 2 non-neural network models based on Logistic Regression and 3 Deep Learning models based on Bidirectional LSTM and used GloVe embeddings, TF-IDF tokenization for feature representation.
• Implemented a linux loadable kernel module which supports allocating memory locations and assign a task to a resource container.
• Executed demand paging for container object mapping to memory & freeing memory on object deletion.
• Implemented a linux loadable kernel module which supported a process container creation, deletion and thread scheduling.
• Locks are used to gaurantee only one process can access a object at a time.
• Designed 4 CNN models namely Resnet-50, Baseline, VGG-16, and Inception, using transfer learing, to perform multi label classification of Amazon satellite images with 17 feature labels like clear, hazy, cloudy, etc.
• Added additional layers and adjusted hyperparameters to increase the efficiency of the baseline model.
• Model then takes an input image, computes a score for each of the 17 features, and then uses a cutoff threshold to decide which labels to be used for classification.
• Coded the Wolf Inns Hotel DBMS application with the user ability to book different category rooms in multiple hotels, check nightly rates of each room along with allowed guests and complimentary services.
• DB admins can access the hotel employee information, and customer details as well as make modifications.
• Developed android application with features of Bill Parsing and Push Notification using Google Cloud Vision API. User can scan bill and get data parsed into the database.
• Uses Node.js for backend where a neural network is designed to predict the expected date for user to run out of gorcery items
• Designed a model to extract features from a geographical image by creating HOG of individual object images and use them for detection.
• Used SVM for classifying and detection of objects in the image within certain range of orientation.
• Detection is done using a sliding window in the image and rescaling the image to detect features from that window to detect what object is present in that image.
• Expertiza is a classroom tool build on Ruby on Rails for group projects and peer review. We refactored assignment model of 'Expertiza' using Ruby On Rails and DRY principles for better architecture.
• Worked on UI issues and improvement of display of review page for students work.
• Added Expert review status and addition of feature of multiple expert review.
• Developed a Car-Rental App with Ruby on Rails for customers to reserve a car, modify reservations, edit check in check out of car, and manage reservations.
• Made use of PostgreSQL for the server and database management and hosted the application in Heroku cloud.
• Developed a Fuzzy Inference System(FIS) based algorithm for simulating the Autonomous Navigation System on MATLAB giving an error of about 1-2% from the ideal path.
• Analysed the developed system against ANFIS and Neural-Networks to prove FIS as the better technique.
• Published on IEEE explore digital library - DOI: 10.1109/ICCCNT.2017.8204164
• Developed an audio input classifier using multiple single hidden layer neural-networks in 5 user defined classes.
• Analyzed extracted audio features using Principal Component Analysis(PCA) and used Neural-Network to learn emotional information from low-level features giving 72% accuracy when tested.