Hi, I'm
Purva
A developer passionate about designing scalable systems, conquering challenges, and staying fit both in code and life! 💻🏋️♂️
Introduction
Overview.
I am a passionate engineer with over 2 years of experience in scalable system design, machine learning, and full-stack development. I specialize in building innovative solutions that drive business success and create real-world impact. My expertise spans data engineering, computer vision, AI, and developing user-friendly applications. With a strong foundation in Python, Java, SQL, and modern frameworks, I am eager to contribute to projects that push the boundaries of technology and foster continuous growth.
What I have done so far
Education / Work Experience
Teaching Assistant
University of Waterloo
- ECE 108
Volunteer Project Developer
UW Blueprint
- Designing and developing a class management platform for Sistema Toronto, enabling 50+ teachers to track absences,upload lesson plans, and claim substitute classes.
- Enhancing teacher workflows with Next.js UI components and backend logic while automating class reminders via Google APIs and cron jobs, reducing missed updates by 40%.
Masters in Computer Engineering
University of Waterloo
- Tools and Methods of Intelligent systems
- Image Processing and Visual Communications
- Data and Knowledge Modelling
- Algorithms Design and Analysis
- Statistical signal processing
- Accounting and Financial Management
- Management and Leadership
Software Development Engineer
Loadshare Networks
- Led migration of legacy codebase to a hexagonal OMS architecture, constituting 65+ APIs reducing build times, API response times, and database size by 35%.
- Built and scaled geocoding APIs, improving address validation accuracy through confidence scoring mechanisms.
- Developed efficient systems and products for cross-utilization of different internal products, increasing orders by 20%.
- Designed and implemented a low-cost data engineering infrastructure, building an end-to-end data pipeline and lakehouse system that improved data analysis efficiency.
- Awarded the Bar Raiser award for demonstrating exceptional dedication and perseverance, exceeding project goals and completing the project under tight deadlines.
Software Engineer Intern
Loadshare Networks
- Developed a Payment Monitoring System, resolving bugs for seamless functionality. Led Rewards module design in the Cohort Engine, boosting user engagement by 30%.
- Played a key role in developing the Reports Framework, enabling efficient data generation and analysis for decision-making.
BTech in Computer Science
MNIT, Jaipur
Tech
Skills.
My work
Projects.
Following projects showcases my skills and experience through real-world examples of my work. It reflects my ability to solve complex problems, work with different technologies, and manage projects effectively.
NasFlow
Developed an automated NAS pipeline for image classification using CNNs and Transformers (e.g., ResNet, DeiT) across diverse datasets, achieving up to 91% validation accuracy with Optuna-based tuning, MLflow/TensorBoard tracking, and GPU-accelerated training.
#python
#MLFLOW
#pytorch
#optuna
#Hugging-Face
ClGanNet(Research Paper)
Developed a novel GAN variant to augment maize leaves, implementing a custom loss function for precise calculations, improving training dataset quality and reducing network size by 40%. Engineered a CNN for maize leaf disease detection, achieving a testing accuracy of 99.04%.
#python
#GAN
#pytorch
Plant-Disease-Detection
Developed a hybrid model for early plant disease identification, combining an Autoencoder and Convolutional Neural Network. Achieved the primary goal of reducing model parameters while maintaining accuracy above 97%.
#python
#autoencoder
#tensorflow
SportsCelebWebApp
Developed a Flask web application that leverages machine learning and image processing to detect sports celebrities in user-provided images. Trained a Random Forest ML model on a diverse dataset, enabling accurate identification of celebrities with confidence scores and names in new uploads.
#python
#flask
#opencv
