Imagination Trumps Knowledge! 

FINCREDIBLE
FEATURE PROJECT

FINCREDIBLE

Designed an ETL pipeline to integrate real-time market ticks and 10K+ daily news articles into MongoDB, powering user-specific feeds and portfolio analytics.

Built gradient-boosted & LSTM predictive models that improved 1-day return-prediction, serving recommendations through a FastAPI micro-service.

Visualized risk metrics and price-impact signals with Plotly dashboards, and deployed the solution on GCP for seamless scalability and faster investor action.

STOCKSAGE
FEATURE PROJECT

STOCKSAGE

Built a Flask micro-service that streams live OHLC data from Alpha Vantage, feeds it to an LSTM (TensorFlow/Keras), and serves rolling price forecasts via REST API.

Visualized actual vs predicted prices and confidence bands with Matplotlib on a responsive dark-theme UI, giving traders an at-a-glance decision aid.

UNIQSCAN
FEATURE PROJECT

UNIQSCAN

Developed a full-stack web platform (FastAPI + MERN) that extracts text with NLTK and OCR, encodes it using a fine-tuned RoBERTa model, and flags plagiarised or LLM-generated content with 93% F1-score, maintaining plagiarism records in a structured MySQL table.

Provided a virtual-classroom dashboard and on-demand visualization reports using Matplotlib, enabling educators to review submissions and uphold academic integrity.

SQL ASSISTANT
FEATURE PROJECT

SQL ASSISTANT

Built a Streamlit tool that converts plain-English questions into parameterised SQL via Google Gemini embeddings, enabling instant data exploration for non-technical users.

Implemented seamless CSV upload with automatic table creation and a live MySQL connection for real-time database querying and results display.

SAFEVIEW
FEATURE PROJECT

SAFEVIEW

Developed SafeView, a real-time browser extension for automated NSFW content moderation using a fine-tuned MobileNet-V3 CNN (F1-score improved from 86% to 92%).

Built a dataset of 500k labeled images, automating data cleaning and augmentation in Python, and trained the model on NVIDIA A100 GPUs.

Engineered a JavaScript + TensorFlow.js pipeline for <50ms latency frame classification, with client-side processing to ensure privacy.

Integrated advanced features like asynchronous buffering, auto-blurring of flagged content, and whitelisting for seamless moderation.

CODESYNC
FEATURE PROJECT

CODESYNC

A real-time collaborative coding platform enabling simultaneous multi-user code editing and execution using Firestore and Dockerized backend.

Integrated a VS Code-style editor with live code sync, session sharing, and remote Python code execution via Flask API in an isolated environment.

PORTFOLIO
FEATURE PROJECT

PORTFOLIO

A portfolio website using Nextjs and TailwindCSS is a great way to showcase skills. It has a clean design and easy navigation with separate pages for projects, experience, skills, and about me.

Projects have details on tools used, skills list expertise, and About Me page has bio, resume, and contact information. It offers a professional online presence for potential clients/employers to learn about skills and accomplishments.