A Master’s student at Dalhousie University who sees software as both art and architecture. Whether it’s designing apps in Flutter, scaling services on AWS, or applying deep learning to computer vision, I focus on making technology both efficient and human-friendly.
Previously at Neurostellar - a BCI company 🧠 and LTIMindtree 🖥, published articles on Medium, co-invented a healthcare robot and national level contest winner, I thrive where curiosity meets execution.
Currently, I’m doing a study on React vs Angular—two frameworks, two philosophies, one goal: building better software.
My calendar is always up-to-date. Please feel free to book a time for a discussion.
Pursuing Applied Computer Science with a focus on Software Engineering and Computer vision with a aggregated GPA of 3.94 / 4.0. Some coursework include:
October 2023 - December 2024 | Flutter, Dart, Python and AWS
Transformed the experience for over 100 users by building a cross-platform application powered by Flutter and AWS, providing mental health insights and improving the overall user experience. Key technical contributions include:
Developed MTooltip, a custom tooltip widget offering complete customization of its shape, position, and content to overcome the limitations of Flutter's default component.
Built a preprocessing module to process JSON strings and remove corrupted data, reducing data loss from Bluetooth sources by over 50% compared to the previous implementation.
Collaborated with cross-functional teams in an Agile environment using platforms like Jira, Confluence, and Github to ensure the timely delivery of a high-quality, bug-free application.
Applied SOLID principles and design patterns, such as Singleton and Observer-subscriber, to improve code quality and maintainability.
Developed impactful developer tools and scalable solutions for the NXP Semiconductors using ARAS PLM, significantly reducing debugging and error-resolution times. Contributed to performance optimization and a seamless infrastructure migration.
Developed a client-side JavaScript debugger for Aras PLM, significantly reducing debugging time and improving developer workflow.
Built a Python utility to automate searches through complex XML configuration files, drastically cutting manual error identification time for the development team.
Collaborated with a cross-functional team of NXP Semiconductors to resolve legacy system issues and deliver key ARAS PLM customizations.
July 2021 - January 2022 | Python, PyTorch and BERT
Earned a one-month internship extension (from 6 to 7 months) in recognition of strong performance and project contributions.
Contributed to a research project proposed by Stanford AI Laboratories by training a Col-BERT model to improve document retrieval accuracy.
Built a Flask application to process user queries and retrieve the most relevant documents using the fine-tuned model.
Public Speaking
Flutter Development at SRMIST
8th March 2024
Invited by SRM University's Cherry+ Network to conduct a hands-on workshop on Flutter Mobile application development. During this "App Archive" session, I guided aspiring developers by leading a live programming demonstration, building the social media application Picspile from the ground up based on Figma designs. This practical, hands-on approach was key to sharing real-world industry insights with a group of highly engaged students.
Developed a full-stack application, Fynd Canada, a platform for booking professional services. Created a Flutter mobile app and a Progressive Web App (PWA) for customers, plus a separate app for professionals to manage their jobs. The platform was built on a serverless AWS Amplify backend and used simple OTP-based login for security. It's key innovation was to build the application in a single codebase - solution published on medium for two different visual designs, making development faster and more efficient.
Recceasy, a cloud productivity suite with its "Storage" mobile app, was developed and released to streamline media management for the signage industry. Through the mobile app, business-critical files can be captured and edited securely on the go, with all data encrypted in transit. The project was built as a tool to solve the unique workflow challenges faced by professionals in the signage and media industry before being deployed.
Dataset Directory
Feb 2021 | Java (Android) and Some Kotlin
Play store Link Dataset Directory, a mobile app with over 100 downloads, was developed and published to streamline the creation of custom image datasets. Through integration with a user's Google account, captured images are automatically stored in Google Drive, and corresponding labels are written directly to Google Sheets via the API. The project was initially built as a specialized tool to support a YOLO object detection research study before being redesigned and released on the Google Play Store.
Academic Projects.
Why this candidate?
Python, PyTorch
Developed and fine-tuned RNN and T5 models for candidate–job matching using the Resume Screening Dataset, achieving a 78% classification accuracy. This result was supported by implementing custom data preprocessing and tokenization pipelines to optimize model training. Further experimentation with selective layer freezing improved the T5 model's performance by an additional 3%, reaching a BLEU score of 0.3 and a BERT score of 83%.
Contributed to the Designite community tool by resolving architecture and code smell issues. PR Merged! Contributed to building the Android Flutter app, Link to Play store Fitbite, by developing its user-facing nutritional tracking chart. Also strengthened the project's backend and deployment pipeline by implementing a Test-Driven Development (TDD) workflow to ensure code reliability, writing a Python script for automated CI/CD analysis using Designite, and containerizing the Spring Boot service with Docker for automated deployments to AWS EC2.
Developed a lightweight, open-source DBMS in Java with MySQL-like CRUD operations. Faster data retrieval and reduced query times were achieved by implementing in-memory processing and a custom data indexing system using HashMaps. The project also included a secure authentication module using MD5 hashing to prevent unauthorized access and a logging system (Console and File) built on SOLID principles for easier debugging.
Developed an open-source Python script using Graphviz to automatically generate Entity-Relationship (ER) Diagrams from textual definitions. This script was built to replace the manual diagramming work required in tools like Lucidchart and draw.io, specifically to accelerate and streamline assignment submissions.
When an application's design calls for different mobile and desktop layouts with identical functionality, the default path can lead to code duplication and complexity. Wondering if there was a better way, I developed a new approach. This article explains how to efficiently build a cross-layout application within a single codebase to avoid this problem. Read More
Go beyond AI to understand image similarity at its mathematical core. This article explains how fundamental image processing techniques can generate values that allow us to calculate a similarity index, providing a clear and effective way to compare two or more images. Read More
Developed and trained a computer vision model using YOLOv-3 and TensorFlow to assess plant health and detect diseases by quantifying chlorophyll levels in leaf images, generating a health score from 0-100. This research was formally published and presented at the 2023 Third International Conference on Artificial Intelligence and Smart Energy (ICAIS), an IEEE-supported conference held in Coimbatore, India.
C. L. K R, et al. "Yolo for Detecting Plant Diseases." ICAIS 2023, doi: 10.1109/ICAIS56108.2023.10073875.Link 🔗
Patent Number: 553295 | Issued on 10/28/2024 Architected the core software for a medical assistance robot, a capstone project, I co-invented to improve elderly care. I programmed the main interaction logic using Python, integrating Jetson Nano and Raspberry Pi 4 hardware to monitor users, manage medication schedules, and automate physician reports. Our work culminated in a prestigious honor at the AICTE Vishwakarma Awards from India's former Union Minister of Education, Shri Dharmendra Pradhan. Co Inventors
When I'm not working, I'm often lost in a book—usually a psychological thriller or a suspenseful crime novel. I particularly like classic authors, especially Agatha Christie, with "And Then There Were None" remaining a favorite. While the list of books that have shaped my perspective is long, I'm currently into "Heal Your Mind, Heal Your Life" by Irfan Alam.
Contact
These are my social profiles. Would be happy to connect with you.
This practical, hands-on approach was key to sharing real-world industry insights with a group of highly engaged students. 
When an application's design calls for different mobile and desktop layouts with identical functionality, the default path can lead to code duplication and complexity. Wondering if there was a better way, I developed a new approach. This article explains how to efficiently build a cross-layout application within a single codebase to avoid this problem. 
Go beyond AI to understand image similarity at its mathematical core. This article explains how fundamental image processing techniques can generate values that allow us to calculate a similarity index, providing a clear and effective way to compare two or more images. 
Architected the core software for a medical assistance robot, a capstone project, I co-invented to improve elderly care. I programmed the main interaction logic using Python, integrating Jetson Nano and Raspberry Pi 4 hardware to monitor users, manage medication schedules, and automate physician reports. Our work culminated in a prestigious honor at the AICTE Vishwakarma Awards from India's former Union Minister of Education, Shri Dharmendra Pradhan. 
