Discover.
Understand the problem, the users and the constraints. Research first — code second.
I am Kaushal Ramoliya — a final-year Computer Science Engineering student and full-stack developer. I build for the web, Android and AI — from React frontends and Node backends to fine-tuned BERT models and CNN classifiers — crafting scalable, efficient solutions that feel as good as they work.

Frontend & Backend — Full-Stack
Android · exploring Flutter cross-platform
BERT fine-tuning, CNNs & predictive analytics
End-to-end web applications — from pixel-perfect, animated frontends to robust APIs and databases. React on the front, Node on the back, shipped fast and built to scale.
Native Android applications in Java, with Flutter in the toolkit for cross-platform builds. Clean architecture, Firebase-powered backends, smooth UX.
Practical, deployed machine learning — fine-tuning Legal-BERT for document classification, training CNNs for image recognition, and building predictive models on real datasets.
Turning raw data into decisions — big-data analysis, visualisation and prediction across e-commerce, streaming and stock-market datasets.
Legal document classification using Legal-BERT fine-tuned on the SCOTUS dataset — 73.86% accuracy across 13 classes. Built with PyTorch and HuggingFace Transformers.
Deep-learning Flask web app that classifies mango varieties (Alphonso, Kesar, Totapuri) with a custom-trained CNN. Built and deployed from scratch for an AI Hackathon.
Big-data analytics suite: e-commerce decision-making, Netflix content analysis and stock-market prediction with Python, Pandas, Plotly and machine learning.
A complete cyber-security lab toolkit — packet analysis with Wireshark, vulnerability scanning with Nmap, forensic recovery, firewall configuration and password auditing.
A compiler that understands Gujarati — if-else constructs written in Gujarati script, lexed, parsed and compiled in C. Language tech meets mother tongue.
Informative website for Navrachana University — programs, schools, admissions, contact and timetables, built with semantic HTML, CSS and JavaScript.
Understand the problem, the users and the constraints. Research first — code second.
Architect the system and sketch the experience. Clean structure beats clever hacks.
Build iteratively with clean, tested code. Tight feedback loops, honest commits.
Ship it, measure it, improve it. A project isn't done until it's live and useful.
GitHub Projects
Technologies
Internships
Years of Code