All Projects

GeoViT is an R&D project tackling urban geolocation using a hybrid AI architecture. Standard geolocation models treat cities as monolithic entities, but Istanbul's 39 districts share similar Ottoman-era architecture, defeating conventional approaches. GeoViT introduces a Context-Aware layer that processes images through a Vision Transformer encoder, extracts text signals via OCR (street signs, municipality markers), detects conflicts between visual and textual predictions, and resolves ambiguity through vector database queries. The system achieves 94.2% district-level accuracy — a 36x improvement over random baseline. The showcase features an interactive terminal demo, Istanbul coverage map with 61,000+ data points, and training data visualizations.

NADCODE is an intelligent chatbot platform designed to transform how businesses handle customer communication. The platform provides AI-powered conversational interfaces that operate around the clock, ensuring customers always receive instant responses. With a simple 5-minute setup process, businesses can embed a fully customizable chat widget on their websites - adjusting colors, positioning, messaging, and behavior to match their brand. The platform includes a comprehensive analytics dashboard for tracking conversations and performance metrics. Built with enterprise-grade security including SSL encryption, data backup, and GDPR compliance, NADCODE supports unlimited concurrent conversations with multi-language capabilities (Turkish and English). The modern tech stack features Next.js 16, React 19, TypeScript, Tailwind CSS, Framer Motion for smooth animations, and Radix UI components.

Unhired is an AI-powered resume optimization platform designed to help job seekers improve their chances of passing Applicant Tracking Systems (ATS) and landing interviews. Users can upload their existing CV in PDF or Word format and input job descriptions via URL or text to receive AI-tailored resume versions optimized for specific positions. The platform features advanced ATS scoring analysis with numerical scores indicating resume compatibility, semantic keyword injection that goes beyond simple matching to understand context and recruiter intent, and format integrity protection that maintains professional formatting while optimizing content. Built with Next.js 16, React 19, TypeScript, and Supabase for backend services, the app leverages Google Gemini AI for intelligent analysis. With rapid 12-second processing times and features like AI cover letter generation and keyword gap analysis, Unhired has helped optimize over 50,000 resumes with an 82% reported interview rate.

Localhost Monitor is a native macOS menubar application designed to solve a common developer pain point: forgetting about running localhost servers during AI-powered development workflows. The app provides real-time monitoring of all active development servers (ports 3000-9000), displaying CPU usage, RAM consumption, and uptime. With smart filtering that only shows development tools (node, python, etc.) and hides system processes, developers can quickly identify and terminate individual processes or kill all servers with a single click. Built with Python using rumps and psutil, it features auto-refresh, a clean section-based UI, and runs entirely from the macOS menubar for seamless integration into your development workflow.