Hello!

My name is Pavitra

About

Hi! My name is Pavitra. I am a 3rd year Software Engineering student at the University of New Brunswick. I have more than 1.5 years of experience doing Software development, Data Engineering and Analysis, and AI at companies and organizations like IBM and Civic Tech Fredericton.

I have research experience with world-renowned researchers like Dr. Rongxing Lu. I am currently doing researching on the Detection of IoT DDoS attacks using Machine Learning algorithms. I am also doing research with Dr. Debasmita Mukherjee on Explainable AI systems.

My hobbies include playing sports like Tennis. I also like singing differnt types of songs. I have made several singing performances at several events at the University of New Brunswick.

My Experiences

Here are some highlights from my journey so far.

IBM - 1 year 4 months

Software Developer Intern

Worked 8 months full-time and 8 months part-time at IBM.
Developed code using languages like Python that works with Whitesource APIs to extract useful information of different PVRs.
Integrated and transformed structured and semi-structured data from diverse sources into delta lake and Synapse Lakehouse, optimizing storage and query performance.
Worked with Python, MATLAB, Node.js, Angular, Flask, Bootstrap, Vue.js, SQL, AQL, APIs, Linux, TypeScript, JavaScript, and Docker.
Improved query performance and storage optimization by applying partitioning, indexing, and caching strategies in Synapse SQL pools and Delta Lake.
Managed and optimized data and analytics environments and platforms, including Azure Data Factory, Azure Data Lake, and Azure Synapse.
Automated data ingestion and transformation workflows, reducing manual intervention and improving delivery timelines.

Kent Building Supplies

Appliance Sales Associate

Sharpened communication skills by performing Sales of atleast $13,000 each month working part-time.
Gained hands-on experience using Microsoft D365, Microsoft Azure Cloud, Microsoft Active Directory, and Microsoft Power BI.

IBM Carbon

Software Developer May 2024 - Present

Developing code for IBM Carbon Open-Source project.
Using HTML, SCSS, Typescript, JavaScript, and Vue.js to develop code.

Civic Tech Fredericton

Software Developer March 2024 - Present

Worked as a Software Developer in the MealPlanner project of Civic Tech Fredericton
Used programming languages like TypeScript, JavaScript, HTML, CSS, Graphql, and sql to develope code.
Developed code using languages like Python that works with Whitesource APIs to extract useful information of different PVRs.
Prepared documentation for the users and Admins of the app.
Performed Data Analysis on various types of data in the app.

UNB - AI and Robotics Research Lab

Research Assistant - May 2024 - Present

Worked as a Research Assistant under Dr. Debasmita Mukherjee
Gained experience with deep learning and robotics.

UNB - AI and Cybersecurity Research

AI Researcher - Thesis Project

Worked under the supervision of Dr. Rongxing Lu (h-index 86) for a research thesis project.
Researched the different ways of detecting DDOS attacks in IoT devices under the supervision of Dr. Lu
Found research gaps in existing and current research.
Evaluated different machine learning algorithms used to detect DDOS attacks in IoT devices.
Worked 8 months full-time and 8 months part-time at IBM.

Lawtons Drugs

Cashier - Part-time - 4 Months

Worked in a team to effectively handle customers, manage cash, and create receipts for customers.
Understood the importance of time management.
Raised customer satisfaction by providing them great treatment using strong communication skill.
Applied various soft-skills while working in a fast-paced environment coordinating and cooperating in a team.

My Projects

Here are my projects.

Help App

Techonologies: Python, Python-Kivy, SQLite, Speech-recognition, Twilio

Created an app with its Frontend in Python-kivy, Backend in Python, and Database in SQLite.
Solved the problem of people not receiving help at the right moment.
Functionality: When a person in need of any type of help says HELP 3 or 4 times in a row, the app recognizes it and sends messages to the persons relative stating that this person needs help at a specific location.
Presented the app at CUSEC 2024 (Canadian University Software Engineering Conference 2024.)
Evaluated by RBC judges.
Received Recongnition as being one of the top 5 best projects submitted to the conference.

Watch on Youtube

Machine Learning - Doordash ETA Prediction

Techonologies: Python, Machine Learning Algorithms, Deep Learning Neural Networks, Scikit-learn, Pandas, Tensorflow

Created a project in a team of 4 to predict the Estimated time of Arrival of Doordash deliveries.
Created Linear, Non-Linear, and Neural Network models.
Linear Models tested on: Linear Regression, Lasso Regression, and Ridge regression.
Non-Linear Models: XGBoost Regressor, Hist Gradient Boosting Regressor, Decision Tree Regressor, Support Vector Regressor, Random Forest Regressor, AdaBoost Regressor
Performed Evaluation and found that XGBoost Regressor Performed the best compared to all of the other models with the lowest 0.642 mean squared error.
Neural Network: Input Layer, 2 hidden Layer, Output Layer

Watch on Youtube

Accident Detection Camera

Technologies: Python, Tensorflow, Keras, deep learning, Flask, SQL, Twilio, Kaggle Datasets.

Road Accidents can happen anytime and anywhere. When road accidents occur, the injured would not be able to get immediate assistance especially when accidents occur in mid nights. This project uses machine learning to train a model that classifies image as accident or not accident. If it is the image of an accident, then the software sends text message along with the location to hospitals and police station using Python Twilio module. Rasberry Pi can be used to make the hardware camera.
Currently training a machine learning model with a large dataset to make the software easily classify images as accident or not an accident.

Watch on Youtube

Carbon Monitoring App

Technologies: Machine Learning, Python, Kivy, SQLite, matplotlib

Participated in the UNB Software Development Hackathon
Created an app with a frontend, backend, and database.
Utilize Matplotlib library in Python to generate a variety of charts showcasing carbon emissions, offsets, and net emissions trends.
Employ Python programming to process user-provided data and execute calculations for determining net carbon emissions.
Implement intelligent decision support systems, potentially incorporating machine learning algorithms, to assist client companies in selecting appropriate partners for carbon offset purchases.
Integrate machine learning algorithms and decision support systems into the application architecture to empower decision-makers with data-driven recommendations.
Ensure an intuitive and user-friendly experience through careful UI/UX design considerations and responsive web design principles.

Watch on Youtube

Covid Face Recognizer

Technologies: OpenCV, dlib, Flask, python mysql connectivity and deep learning

A website made with Python-Flask integrated with a Face Recognition Software written in Python connected to MySql Database
Worked on a wide-variety of software developemnt tasks.
Logic: When a person becomes COVID positive, Doctors can log into this website and enter the persons unique identification number and turn the persons status to covid Positive. The database of this website is connected to a Face-recognition software written in python. The database has a relation with the persons face encodings, name, unique identification number, and the current covid status. Now, the persons covid states is changed to positive. The camera uses this covid classification software that is connected to the database. So, now when a person enters a shop, their face encodings would be scanned and searched for in the database. Once the face encodings are found, the software will check if the persons covid status is positive or negative. If positive, the shopkeeper would notice a red box around the face of the person making the shopkeeper know that the person is covid positive.

Rover - Embedded System Design course

Technologies: C Programming Language, Embedded System Design, Robotics

Worked in a team of 4 to build a rover controlled by a transmitter that detects magnetic anomalies using a hall effect sensor.
Won 2nd place in the Harvest Robotics Competition organized by the UNB.
Developed embedded system in C and C++ programming languages that controlled the transmitting and receiving actions of the embedded system.

Watch on Youtube

Portfolio Website

Technologies: Next.js, React.js, HTML, Tailwind CSS, JavaScript

Created a portfolio website using Next.js, React.js, HTML, Tailwind CSS, JavaScript.
Effectively utilized the fundamentals of Software Engineering, Software Development, and Cybersecurity to create this website.

Face Detection Camera for security

Technologies: OpenCV, dlib, Flask, python mysql connectivity and deep learning

Made a face detection camera that checks if the face encodings scanned are a part of the MySql database. If the face encodings of the person are a part of the MySql database,then the person is an employee of the company and thus would be allowed to go inside the building. However, if the persons face encodings would not be found in the MySql database, then the persons face would be highlighted with a red box and a beep sound will ring.

Watch on Vimeo

Book Cricket Game

Technologies: Python

Made a Book Cricket game using python and mainly its Random module.
Logic: The last digit of each page of a book is equal to the number of runs scored by a player in each ball. If the last digit exceeds 6, then the runs start from 1, 2 and so on. If the last digit is zero, then the player is out. In the end of the game, the team that won would be displayed along with the player who scored the maximum number of runs. The random module was used to generate the random page number.
Got a full score for this Grade 11 final project.

Uniform Detection Model

Technologies: Technologies: Python, Tensorflow, Keras, deep learning, Flask, SQL, Twilio, Kaggle Datasets.

Tried to train a Uniform Detection Machine Learning model that classifies students based on whether they are in proper school uniform or not.
Faced the challenge of gathering a big dataset. I clicked some photos of me to train the model. However, the dataset turned out to be extremely less. I tried augmentating the images to increase the dataset. But it did not prove to be a much larger dataset.