I am Software Engineer & Full-Stack Developer who built end-to-end web platforms and AI-powered systems with a strong focus on system design, database architecture, and reliable APIs. Across the stack, I build responsive interfaces, design scalable backend services, and develop practical machine learning pipelines. I also integrate modern LLM APIs directly into production workflows, turning complex engineering decisions into reliable, real-world software.

Languages & Frameworks:
JavaScript (ES6+) • React.js • Node.js • Express.js • Python • Flask • HTML5/CSS3

Databases & Tools:
MongoDB • RESTful APIs • Git & GitHub • GitHub Actions • VS Code • Postman

A full-stack blogging platform built solo to practice real-world authentication, database design, and testing patterns—not a tutorial clone. Deployed live on Vercel with automated CI/CD.

• Backend Architecture: Built REST APIs with Node.js and Express.js, backed by MongoDB across 9 collections via Mongoose.
• Authentication & Security: Implemented dual-token JWT auth (15-min access token, 7-day rotating refresh token) with bcrypt password hashing.
• Database Optimization: Added compound indexes in MongoDB to speed up feed queries and prevent duplicate records.
• Analytics: Built a view-tracking feature using IntersectionObserver to log genuine, verified article reads.
• Frontend & State: Integrated TanStack Query on React for smart background caching and optimistic UI updates.
• Automated Testing: Wrote unit and integration tests (Jest, Supertest, Vitest) covering auth lifecycles, post CRUD operations, and core endpoints.
• Links: Live Demo • GitHub Codebase

Final-year capstone project: An image classification system that identifies 7 categories of skin disease from photos, trained on the HAM10000 dataset (10,000+ images). Combined three pre-trained CNN models into an ensemble reaching ~87% test accuracy.

• Deep Learning Ensemble: Built an ensemble of 3 CNN models (EfficientNetB3, ResNet101, DenseNet121) in TensorFlow/Keras, combining predictions via softmax averaging.
• Computer Vision Pipeline: Built an OpenCV preprocessing pipeline for image resizing (224×224), normalization, and data augmentation to handle dataset class imbalance.
• Backend & Authentication: Built a Flask REST API backend with JWT-based authentication for secure image uploads and prediction endpoints.
• Database & Logging: Logged prediction metadata, diagnosis histories, and inference metrics to MongoDB using PyMongo.
• Frontend Dashboard: Built a React 18 frontend featuring drag-and-drop image uploads and an interactive diagnosis results dashboard.
• Ownership: Worked in a 5-person team; directly owned and engineered the ML pipeline, backend API, and frontend dashboard.
• Links: Live Demo • GitHub Codebase

• LinkedIn: Purvesh Jathade
• Email: Purvesh9997@gmail.com