Selected Projects

• Designed and deployed sophisticated multi-agent AI assistant using LangGraph orchestration with 6+ specialized agents including Security Agent , Security Explanation Agent, Intent Classifier, and Persona Manager for contextual threat detection and conversation management.
• Engineered multi-source knowledge base blending structured data, personal project lore, and adaptive personas to simulate natural dialogue across professional, technical, and personal domains with comprehensive security-first architecture.
• Implemented advanced multi-agent reasoning system using custom Retrieval-Augmented Generation (RAG) pipelines, intent classification, tone modulation, and persona switching with multi-model fallback system and automatic quota management.
• Built sophisticated threat modeling system with multi-layered defense: input sanitization, AI-powered threat detection, and contextual security responses that successfully block prompt injection, system extraction, and social engineering while preserving natural dialogue flow.
• Developed novel Security Explanation Agent that generates contextual, personality-consistent responses for blocked content instead of robotic security messages, enhancing UX while maintaining comprehensive security.
• Built secure backend infrastructure using FastAPI (Python), Gunicorn, Nginx reverse proxy, and Cloudflare Workers for API gateway isolation with session-based rate limiting and real-time monitoring.
• Deployed full-stack system on self-managed Hetzner VPS, handling bare-metal Linux setup, SSL provisioning via Let’s Encrypt, server hardening, key management, and backend service orchestration.
• Integrated Cloudflare DNS and proxying to shield origin server, separate public API keys from client-side code, and securely serve the application globally with conversation persistence using SQLite.
• Managed deployment flows through direct server-side Git integration and shell-based service reloads with frontend hosted via GitHub Pages connecting dynamically to backend APIs through serverless Cloudflare Worker routing.
• Full ownership across AI system architecture, security engineering, conversational UX design, multi-layer backend security, server provisioning, deployment orchestration, and API integration.

• Developed AI-powered debate system to refine user reasoning through structured counterarguments.
• Built Reddit-style platform enabling structured debates with dynamically generated AI challenges.
• Integrated multi-agent AI reasoning to produce nuanced counterperspectives.
• Designed adaptive learning system delivering real-time feedback based on user performance.
View Source Code →

• Developed interactive world map visualization with D3.js and Python for exploratory data analysis.
• Enabled pan/zoom, hover tooltips, and line chart popups for enhanced UX.
• Visualized country metrics including GDP, energy production, and happiness index.
View Source Code →

• Built C-based cache simulator analyzing miss rates, associativity models, and bus snoop traffic.
• Supported varying coherence protocols and block size dynamics.
View Source Code →

• Developed terminal-based virtual pet simulation using functional programming paradigms.
• Architected fully recursive game logic; incorporated concurrency via Lwt library.
View Source Code →

• Designed and deployed a responsive static site using HTML, CSS, and JS.
• Focused on speed, information hierarchy, and minimalistic UI principles.
View Source Code →