Work, in practice.

Applied AI, research, and software built around real problems. Explore the context, approach, and outcome behind each project.

FreightCollab

A comprehensive customer relationship management system built for CircleBit that revolutionized freight logistics operations. The system replaced a legacy platform with modern, performant architecture.

Delivered a system that was 17× faster than the legacy platform, featuring 20+ interactive dashboards and ML-powered forecasting capabilities.

  • React
  • D3.js
  • GCP
  • BigQueryML
  • JavaScript
  • Node.js
  • Data Visualization
Read the project details

The problem

CircleBit's freight logistics operations relied on OFreight, a legacy system that was slow, unresponsive, and couldn't handle the growing data demands of modern logistics.

The approach

Designed and built a modern, full-stack CRM from scratch using React for the frontend, integrated with GCP services for backend processing, and BigQueryML for predictive analytics.

Impact

  • 17× performance improvement over legacy system
  • 20+ interactive dashboards with D3.js visualizations
  • ML-powered forecasting with BigQueryML
  • Real-time decision making capabilities
  • Transformed freight logistics operations

API Test Automation

Intelligent testing system that uses machine learning to automatically generate comprehensive API test cases and provides automated grading for validation.

Achieved 80% improvement in test coverage while dramatically reducing manual testing overhead and improving overall code quality.

  • Python
  • TensorFlow
  • Next.js
  • Vercel
  • API Testing
  • Machine Learning
  • Automation
Read the project details

The problem

Manual API testing was time-consuming, inconsistent, and couldn't keep up with rapid development cycles, leading to poor test coverage and quality issues.

The approach

Developed an ML-powered system using TensorFlow that analyzes API specifications and automatically generates comprehensive test cases, with automated grading pipeline for validation.

Impact

  • +80% improvement in API test coverage
  • Automated test case generation using ML
  • Intelligent grading and validation pipeline
  • Reduced manual testing overhead
  • Improved code quality and reliability

Dubai World Challenge

Local Academia Category winning project featuring a semi-autonomous transportation system with augmented reality interface for passenger interaction.

Won the Local Academia Category at Dubai World Challenge, demonstrating innovative integration of autonomous systems with augmented reality.

  • Computer Vision
  • AR
  • IoT
  • Robotics
  • Sensors
  • Real-time Systems
Read the project details

The problem

Traditional transportation systems lack intelligent automation and intuitive passenger interfaces, limiting efficiency and user experience.

The approach

Developed a semi-autonomous system combining computer vision, IoT sensors, and AR technology to create an intelligent transportation solution with immersive passenger experience.

Impact

  • Dubai World Challenge - Local Academia Category Winner
  • Semi-autonomous system with intelligent navigation
  • AR interface for enhanced passenger experience
  • Integration of computer vision and IoT technologies
  • Demonstration of practical autonomous systems

WiFi Domain Generalization

Research project investigating domain generalization techniques for WiFi sensing systems, focusing on maintaining performance across different environmental conditions.

Published research in IEEE CAIT 2024 and ongoing thesis work extending the framework for comprehensive real-world deployment.

  • Deep Learning
  • Domain Adaptation
  • Signal Processing
  • WiFi
  • Neural Networks
Read the project details

The problem

WiFi sensing systems struggle to maintain performance when deployed in environments different from their training conditions, limiting practical applicability.

The approach

Developed novel domain adaptation techniques that allow WiFi sensing models to generalize across different environments while maintaining computational efficiency.

Impact

  • Novel domain adaptation framework for WiFi sensing
  • Cross-environment robustness improvements
  • Computational efficiency maintained
  • Real-world validation across multiple scenarios
  • Published in IEEE CAIT 2024

Collabrain

A real-time collaboration and brainstorming platform supporting teams, social connections, file sharing, multiplayer whiteboards and documents, video calls, multiple languages, and text-to-speech.

Delivered a multilingual collaborative workspace with synchronous creation tools and text-to-speech support.

  • Next.js
  • Redis
  • Node.js
  • Firebase
  • Oracle Cloud Storage
Read the project details

The problem

Collaborative work is often fragmented across separate communication, file, meeting, and ideation tools.

The approach

Combined team collaboration, messaging, video, shared documents, whiteboards, and file storage in one web platform using Next.js, Redis, Firebase, Node.js, and Oracle Cloud Storage.

Impact

  • Multiplayer whiteboards and documents
  • Team, friend, and messaging workflows
  • Video calling and file sharing
  • Multilingual and text-to-speech support

ResearchHubble

A NASA Space Apps Challenge project for open-source and open-science collaboration, using CV parsing to identify skills and recommend relevant projects.

Created a project-discovery experience that connects contributors with relevant open-science opportunities.

  • React
  • Node.js
  • Firebase
  • Recommendation Systems
  • Semantic Search
  • Fuzzy Search
Read the project details

The problem

Potential research contributors struggle to discover open projects that match their skills and interests.

The approach

Combined CV parsing, recommender systems, semantic retrieval, and fuzzy search in a React, Node.js, and Firebase platform.

Impact

  • Automated skill extraction from CVs
  • Personalized open-project recommendations
  • Semantic and fuzzy project search
  • Designed for open-source collaboration

Terminal Portfolio

A modern, interactive portfolio website featuring a terminal-style interface, typing game, and comprehensive showcase of projects and research.

Created an engaging, keyboard-first portfolio that demonstrates technical skills while providing comprehensive project and research showcase.

  • Astro
  • React
  • TypeScript
  • Tailwind CSS
  • Performance Optimization
Read the project details

The problem

Traditional portfolios are static and boring, failing to showcase technical skills and personality effectively.

The approach

Built an interactive terminal-style portfolio using Astro and React, featuring command-line navigation, typing game, and performance optimization.

Impact

  • Interactive terminal interface with command system
  • Typing game with AI-themed text corpus
  • Performance optimized (LCP <2.0s)
  • Comprehensive project and research showcase
  • Demonstrates technical skills through implementation