Master of Engineering from University of Toronto 2023 | COEP ‘21
Check out my plum profile:
Electrical And Computer Engineering
Specialization - Analytics
Graduation : December 2023
GPA : 3.92/4.0
Coursework: Data Analytics | Reinforcement Learning | Big Data Science | AI in Finance | Blockchain & Cryptocurrencies | Digital Image Processing | Machine Learning | Design of Intelligent Sensor Networks | AI Development in Healthcare | Cloud Computing
Major - Instrumentation And Control Engineering
Minor - Computer Science
Graduation: June 2021
GPA – 8.85 / 10
Coursework: Object Oriented Programming & Development | Data Structures, Files & Algorithms | Artificial Intelligence & Machine learning | Database Management Systems | Automatic Control Systems & Design | Microcontrollers and Embedded Systems |Digital Signal Processing |Power Electronics |Analog and Digital Electronics
Led critical contributions to the development of comprehensive use cases for the IBDT project, driving the creation of a dynamic web portal for analyzing data obtained from sensors in different buildings on UofT campus to gain deep insights about energy usage and carbon footprint generated by each building. Engineered essential core functionalities within the portal, elevating user experience by facilitating seamless browsing of datasets and buildings, advanced filtering, and group download capabilities, using cutting-edge technologies such as React, Next.js, and TypeScript. Implemented CI/CD pipelines with GitLab to automate the build and deployment process, containerized applications using Docker for streamlined development and environment consistency.
Spearheaded the development and testing of a machine learning model for emotion recognition on images using a deep learning model as part of a company project to improve the culling of professional photographs. Leveraged machine learning tools to develop and optimize the data processing pipeline by using expertise in Python, PyTorch, TensorFlow and Keras along with feature extraction from images and conversion to a format suitable for the deep learning model. Successfully deployed the machine learning pipeline after unit testing and code reviews on AWS, resulting in improved accuracy and speed of the image culling process and overall project efficiency.
Developed a web-service using the Django framework. This involved designing and implementing APIs to fetch data from a database, as well as performing data cleaning and validation. I also generated dashboards on the front-end for data visualization and streamlined the data retrieval and reporting process to improve organizational efficiency. Throughout the internship, I gained valuable experience in version control using Git and working in an Agile development environment.
Designed a Luenberger observer using MATLAB/Simulink and state-space theory to estimate current in Li-ion batteries. Classical control algorithms were applied for battery current estimation, and analog simulations of the automated system were carried out on Proteus and Multisim. The implementation of the Luenberger observer improved the overall accuracy of the current estimation process, leading to an increase in efficiency and reliability of the battery monitoring system.
The current workflow of a TalkMaster includes answering intercom calls from classrooms, resolving the issues remotely using various apps, and dispatching staff to room if necessary. This process is time-consuming and requires experienced personnel to provide support. To address this, we leveraged the capabilities of large language models and built an AI solution to aid in decision-making.
The escalating global concern surrounding mental health issues has been accentuated by a
myriad of stressors originating from contemporary societal dynamics.
Our approach encompasses addressing the stigma and barriers associated with traditional mental
health care, acknowledging the reluctance of individuals to seek professional help due to fear of
judgment or vulnerability.
The final product is an AI-powered Mental Health Chatbot that delivered accessible, anonymous, and stigma-free mental health support. This chatbot employed AI-driven emotional support, offered guidance on self-help strategies, and, when necessary, facilitated connections to professional help. A key differentiator amongst our competitors was the facility to assess the user's distress level, provide tailored solutions, such as recommending remedies, connecting with helplines, and alerting loved ones. By developing a mental health solution, we took a holistic approach to mitigate the psychological toll of modern-day stressors and supported mental well-being across diverse demographics.
Presentation:
Designed a controller for predicting the controller command for ailerons using the airplane's status. Compared the results of a linear regression model and all gradient descent models. Finally fine tuned a mini-batch gradient descent model to give the best results with low validation and test RMSE (root mean squared error)
We created a Cloud Lab Inventory management project that allows for the electronic storage and management of device information, making it easier for both users and Lab Inventory administrators to track and locate devices. This eliminates the need for manual work and reduces the chances of errors. The system consists of two modules, one for administrators and one for users, who can register, issue, rent and renew devices, and contact the Lab Inventory team for any queries. AWS services were used to develop this project, including AWS Lambda, Zappa, S3, DynamoDB, and SES
Cloud Computing
A web front end that allows users to upload a new key and image, uniquely identifying the image. The system stores the image in the local file system and adds the key to the database. Users can update or delete the stored key and image. A page displays all available keys and images, with a button to clear the application of all keys and values. Additionally, there is an option to configure and display statistics for the mem-cache, including capacity, replacement policy, and cache clearance. The application consists of four components implemented as a separate Flask instance-
Intelligent Sensor Networks
The main objective of the project was to develop a low-level and user-friendly
system that resembled the functionalities of an off-the-shelf gas monitoring system.
The primary focus was to reduce the cost of the developed system as compared to
the typical modules (e.g., Honeywell’s E3 Point)
Report
Demonstration:
Digital Image Processing
The goal of this project was to study the tools and technologies required to transfer knowledge
from a larger model to a smaller one that can be used practically in real-world settings.The
project focused on the setting of Knowledge distillation as a model compression technique.
Project Repo: https://github.com/RuchitaBhadre/Knowledge-Distillation-for-Building-Lightweight-Deep-Learning-Models
Blockchain & Cryptocurrencies
The Pet-Shop DApp project is a blockchain-based application that has been developed by utilizing tutorials such as pet shop and elections. Used Ganache to fire up an Ethereum blockchain and connected it with MetaMask. Truffle was used to develop the DApp and write smart contracts for sending ether, voting for favorite pet and adopting a pet along with frontend enhancements like enhanced UI. The project includes all the functionalities of the tutorials with additional features incorporated into the pet shop template. Once the user enters the pet store, they can choose and adopt a dog of their choice and pay for it. The project also includes a donation feature where the user can donate Ether to the store. The user's account information is displayed on the welcome message, and the Pet Shop Status panel shows the number of clients in the shop, the total number of pets adopted, and the number of pets the user has adopted. Additionally, users can vote for their favorite pet, and the total number of votes each pet has received is displayed on the webpage. The UI of the project has been modified to include calming colors, a header, and a navigation bar for easy access to the donation portal and main shop page.
Project repo:https://github.com/RuchitaBhadre/Pet_shop-DApp
AI in Finance
Built 3 neural networks in Tensorflow and Keras that predict stock market prices using daily stock data. The model with RNN and LSTM cells performed the best in time series forecasting. The results were evaluated with financial metrics and white reality check.
Project Repo:https://github.com/RuchitaBhadre/Stock-Price-Prediction
Cloud-based Data Analytics
Constructed a predictive model that uses the physiochemical characteristics to determine the wine's quality. We also tested the data with multiple models using AutoML and Hyperparameter tuning in Microsoft Azure.
Project repo: https://github.com/RuchitaBhadre/wine-quality-prediction.git
Automation
Created an automated mask for COVID-19 in a team of three which detects other people coming within a 1 meter radius of the mask-wearing individual using infrared and motion detection processed and implemented using machine learning on an Arduino microcontroller.
Associated paper: Bhadre et al., (2021) SmartMask- Developing an automated self-care system https://arxiv.org/abs/2207.01492
Demonstration:
Internet of Things
Constructed an IoT-based agricultural robot, AGRIBOT, in a team of two with a sensing and data analytics system for monitoring and controlling the microclimate in large-scale greenhouses to mitigate labour costs while increasing productivity which was presented at the national level ‘Eureka Hackathon 3.0’.
Associated paper: Bhadre & Yeole (2020). Deployment of AGRI-BOT in Greenhouse Administration http://arxiv.org/abs/2206.07266
Final product:
Web Development
Created a browser extension in a team of six using open-source tools to make websites compatible for people with visual impairment, auditory impairment, dyslexia, or motor disabilities, and presented at the ‘Smart India Hackathon’ to represent COEP.
Object Oriented Programming and Development
Developed a database management system for an entertainment agency using OOP to ease file handling by creating modules for different actions like event management, revenue management which simplified the job of the manager.
Computational Theory
Investigated different flow regimes (laminar/turbulent flow) by simulating fluid flow behaviour in MATLAB/Simulink to calculate Reynold’s number of a fluid, and estimate likelihood of solid deposition in the pipe for viscous fluids.
Tech2U is a program offered by LSM to humanize tech support across UofT campus. Responsibilities include:
Completing administrative tasks such as organizing confidential documents, assisting students with sign-up procedures, escorting students, and managing exam rooms.
