Hi, I'm Suman

Web Developer | Web3 & Cybersecurity Enthusiast

Passionate web developer with a strong grasp of full-stack technologies, cybersecurity, and UI/UX. Exploring Web3, smart contracts, and building secure, scalable digital solutions.

Skills

What I Can Do

Machine Learning

Python

React

Artificial Intelligence

JavaScript

R

SQL

Django

Full Stack Development

Github

Docker

MongoDB

NodeJS

C

Web3.js

Ethereum

Portfolio

What I've Built

Here are some of my recent projects showcasing different technologies and skills

Allocate Duty and Press Note Management
Web Dev

Allocate Duty and Press Note Management

Developed a full-scale Police Management System in ASP.NET with modules for Press Note and Duty Management, automating duty allocaton/revision, email notifications, and database-driven workflows. Optimized UI/DB design imporved scalability, real-time communication, and reduces manual intervention by 90%

ASP.NETSQL ServerBootstrap
Webpage for the MEP Kolkata-based Company
Web Dev

Webpage for the MEP Kolkata-based Company

Developed and deployed pauldeltaarc.com using Next.Js with responsive, SEO-Optimized design, ensuring 99.9% uptime and cross-platform accessibility. Iteratively enhanced features based on analytics and feedback, boosting performance and user satisfaction.

NextJSTailwind CSSVercel
MemoMind - Full Stack Link Management Application
Web Dev

MemoMind - Full Stack Link Management Application

Built a full-stack management app wiht secure JWT authentication, public mind sharing, and advanced tagging/categorization. Optimized frontend for a perfect Lighthouse score (100/100) with sub-second load times, and planned AI integration for querying YouTube, documents, and X content. Future scope includes AI agents to enable users to interact with past links and generate insights.

Next.jsTypeScriptRenderVercel
Quora-Question-Pair-Similarity
ML/AI

Quora-Question-Pair-Similarity

Quora Question Pair Similarity – Built an ML solution to detect duplicate Quora questions using Logistic Regression, SVM, and XGBoost. Achieved best performance with XGBoost (accuracy ↑, log loss 0.35%), improving search, recommendations, and user experience.

PythonScikit-learnXGBoostPandasNumPyNLTK
Matches-Celebrity-Faces
ML/AI

Matches-Celebrity-Faces

Developed a deployable face detection and recognition system using the pre-trained VGG-Face CNN model, leveraging cosine similarity for accurate celebrity identification, optimized large-scale feature matching, and applied deep learning techniques to real-world use cases in entertainment, security, and personalized recommendations.

TensorflowKerasOpenCVNumPy