Building machine learning systems and full-stack applications — from EMA/RSI trading algorithms to LSTM emotion classifiers.
B.Tech Computer Science Engineering (AI & ML) student at Techno India University, Kolkata. I like taking real-world problems — market signals, interview prep, splitting the dinner bill — and turning them into working systems end to end.
I'm a Computer Science Engineering student specializing in Artificial Intelligence and Machine Learning, currently in my third year at Techno India University, Kolkata. My work sits at the intersection of applied ML and full-stack engineering — I'm equally comfortable tuning a classification model as I am shipping the React dashboard that serves it.
Recent projects range from an algorithmic trading platform with live broker execution, to an AI interview coach powered by the Gemini API with real-time speech recognition, to an NLP pipeline that classifies emotion from raw text using LSTMs. I've also completed virtual internships with IBM, Google for Developers, and Fortinet, and simulations with J.P. Morgan, Goldman Sachs, Accenture, and Electronic Arts via Forage.
I'm looking for opportunities to apply AI/ML and full-stack skills to problems with real users and real stakes.
Full-stack algorithmic trading platform for NSE/BSE equities implementing an EMA(9)/EMA(21) crossover strategy with RSI(14) confirmation and ATR-based position sizing. Live broker execution via Zerodha Kite Connect and Upstox OAuth, with a React dashboard for backtesting, live signals, and portfolio tracking.
Full-stack AI-powered interview prep platform with resume ATS scoring, Gemini-generated mock interview questions, and voice-based interviews with real-time speech recognition. Includes an evaluation engine scoring technical skill, communication, and confidence, plus a personalized career-advisor module.
NLP-based system classifying human emotions from text using ML and deep learning models (LSTM / neural networks). Applied tokenization and TF-IDF / embedding preprocessing, evaluated with accuracy, precision, recall, and F1-score.
Full-stack expense-splitting app supporting equal, exact, percentage, and share-based splits with multi-payer and direct (non-group) expenses. Implements a debt-simplification algorithm that reduces group balances to the minimum number of settle-up transactions.
Adaptive data-science platform for industrial analytics — takes raw industrial data through to a deployed, explainable ML model. FastAPI service layer, React frontend, containerized with Docker, backed by an automated test suite.
Full-stack Library Management System built as a university software engineering project, with role-based workflows for Admin, Librarian, and Member users covering catalog, circulation, and member management.
Bilingual, Hindi-first financial literacy and planning assistant for rural households — computes a financial health score, gives priority-ranked savings and budgeting advice by occupation, and helps users discover relevant government schemes, all designed for low-end devices and slow connections.
Edge-deployed full-stack app that predicts student academic performance from a multi-factor scoring model, with an interactive analytics dashboard for educators to make data-driven decisions. Built on Cloudflare's edge runtime with a distributed SQLite database.
Full-stack event discovery app where users create, browse, and RSVP to events with live attendee counts and smart filtering by category, keyword, and date. Responsive across desktop, tablet, and mobile.
A front-end recreation of the Amazon India shopping experience — responsive layout, interactive product browsing, and UI animations built from scratch with vanilla HTML, CSS, and JavaScript.
More on github.com/Rupam179