Blog & Articles
Deep dives on the systems I build — retrieval pipelines, deployment tooling, and performance work.
Building 70+ Custom Gutenberg Blocks for a NYC Real Estate Brokerage
At SLTech I built BHS Custom Blocks, a WordPress plugin with 70+ Gutenberg blocks for Brown Harris Stevens, one of New York's best-known residential brokerages. Here's how a block library that size stays easy for editors and easy to maintain.
AI / ModelsGemini 4 Argon Wins 14 of 19 Benchmarks. The Other 5 Matter More.
Google's launch table puts Gemini 4 Argon ahead of GPT-6 Astra, Claude Fable 5.1 and Claude Opus 5.5 on most benchmarks. Here's how I read it as someone who builds production AI features, and why the rows where it loses are the interesting ones.
AI / RAG SystemsBuilding a Production RAG Chatbot with OpenAI, Chroma, and Claude
How I built a knowledge-base chatbot for REConnect — from chunking source docs and retrieval with Chroma Cloud to agentic tools, rolling summaries, and thumbs-down retry logic.
DevOps / MCPPrompt-Driven Deployments: Bridging Claude MCP with Coolify via an OAuth Shim
Coolify's API uses a static bearer token; Claude's remote MCP connectors expect OAuth. Here's the small Node/Express shim I built to connect them — with Supabase-backed sessions, rate limiting, and a consent step.
PerformanceCutting API Response Times by 40% with Redis, Query Tuning, and Lazy Loading
A practical breakdown of the caching strategies, query optimisation, and lazy loading I used at The Genius Group to reduce API response times by roughly 40% under production load.