Selected Projects

Backend systems, AI platforms, and practical tools.

These projects emphasize clean APIs, modular services, testing, deployment, and applied machine learning.

HeuriGuard web app preview on laptop

2025 - 2026

HeuriGuard - AI Cybersecurity Platform

HeuriGuard is a multi-module threat detection platform for malicious URLs, phishing links, and SMS spam classification. I led backend development, implemented modular REST APIs, engineered feature pipelines, optimized processing for latency and reliability, and maintained deployment scripts for Docker-based releases.

PythonFlaskMLNLPAndroidRenderDocker
  • Hybrid rule-based and machine learning detection for URL and SMS safety.
  • Feature extraction for URL structure, domain characteristics, and text signals.
  • Production deployment with a custom domain and real-time predictions.
RAG assistant architecture diagram

2026

RAG-Based Customer Support Assistant

A retrieval-augmented assistant focused on reliable document ingestion, retrieval, routing, fallback handling, and graceful degradation for edge cases. The project strengthened my work with agentic AI workflows and maintainable AI application architecture.

PythonLangGraphGoogle GeminiChromaDBRAG
  • Modular document ingestion and retrieval pipeline.
  • Routing and fallback mechanisms for uncertain queries.
  • Maintainable design for production-style customer support flows.
Swagger UI for FastAPI food delivery backend

Backend API

FastAPI Food Delivery Backend

A real-world backend application that manages menu operations, customer orders, cart workflows, and checkout processing. It simulates common food delivery platform features while keeping the API clean, validated, and easy to test.

PythonFastAPIPydanticUvicornREST APIsSwagger
  • Menu CRUD, cart updates, checkout workflow, and order processing.
  • Case-insensitive search, sorting, pagination, and combined browse endpoint.
  • Pydantic validation and Swagger UI API testing.
Malicious URL detection system screenshot

Machine Learning

Malicious URL Detection Web Application

A deployed web application that detects malicious URLs with 95%+ accuracy using machine learning classifiers and a feature extraction pipeline for URL patterns, domain characteristics, and behavioral indicators.

PythonFlaskScikit-learnREST APIRender
  • Real-time URL safety verification through a web interface.
  • Detailed classification reports for safer browsing decisions.
  • Cloud deployment for public access and practical use.
Text-to-image generation project preview

IGNITE HACK 2024

Text-to-Image Generation Using Stable Diffusion

A generative AI web application that converts text prompts into high-quality images. The generation pipeline was optimized with efficient preprocessing and caching, reducing latency by 40%.

PythonStable DiffusionGenerative AITensorFlowGoogle Colab
  • Secured 3rd place at IGNITE HACK 2024 among 50+ teams.
  • Focused on prompt-to-image quality and generation speed.
  • Showcased applied deep learning in a creative workflow.

More Projects

Additional engineering practice.

Image Classification Using Deep Learning

CNN project with data augmentation, transfer learning, model training, evaluation, and hyperparameter tuning.

Online Mock Test Application

Java Swing application with automated grading, real-time analytics, timer functionality, and a question database.

YouTube Video Downloader

Python Tkinter desktop application supporting multi-quality downloads with progress tracking and error handling.