Projects
A selection of AI architecture work and technical explorations I've built — from enterprise RAG systems to developer tool comparisons.
Research & Concepts
Context Graph for AI Agents
Exploring how context graphs give AI agents structured memory, relationships, and reasoning. Covers knowledge representation, graph construction, context retrieval, agent reasoning, persistence, and evaluation — a mind map of the key concepts behind intelligent context management.
Virtual Assistant with Knowledge Graph
A next-generation virtual assistant powered by a knowledge graph to deliver highly accurate and context-rich product information. Uses natural language processing and structured data relationships to provide precise answers about product features, specifications, and availability.
Featured Projects
Agentic Retrieval Augmented Generation
A 6-layer architecture deep dive into how autonomous agents orchestrate retrieval, reasoning, and tool use to deliver grounded, accurate answers. From data sources and indexing through hybrid retrieval, tool use, agent reasoning, and response synthesis with fact verification.
My MCP Journey
From solving one problem for our RAG agents to building an enterprise-grade MCP Gateway — a 5-phase journey with no roadmap, just solving the next bottleneck. Covers local MCP servers, remote transport via SSE, OAuth with Microsoft Entra ID, multi-server gateway, and Claude Skills integration.
Learning Claude Skills
A step-by-step journey into extending Claude Code with custom capabilities — from discovering slash commands to building production-ready Skills. Covers SKILL.md authoring, prompt engineering, MCP server integration, hooks, and multi-tool workflows.
Claude Code vs Kiro CLI
A comprehensive comparison of two AI-powered coding assistants with different philosophies. Claude Code offers agentic freedom with MCP servers and Skills. Kiro provides spec-driven, structured development with guardrails. Includes side-by-side workflow analysis.