Blog

Engineering notes from real builds

AI
From LangGraph to Gemini Enterprise Agent Platform: the LangGraph logo joined by a line of light to the Gemini Enterprise logo

From LangGraph to Gemini Enterprise Agent Platform

Rebuilding a LangGraph and pgvector RAG agent on Agent Retrieval, ADK and Agent Sessions. What mapped cleanly and what had to change.

Read article →
Building adaptive RAG on Google Cloud: hybrid retrieval, RRF fusion and a two-node LangGraph controller

Building Adaptive RAG on Google Cloud

Chunking, embeddings, hybrid search, reranking, and a LangGraph agent that decides when to search again. How each layer was built, and what we changed after measuring it.

Read article →
Vancouver Sommelier AI, a multi-agent drinks concierge

Building a Multi-Agent AI Sommelier for Vancouver

A multi-agent system that searches live inventory over MCP, recommends pairings, designs beverage menus from a food menu photo, and measures its own hallucination rate.

Read article →
SUM AI Voice, a real-time voice assistant on the website

Building SUM AI Voice: a Real-Time Voice Assistant on the Gemini Live API

Press one button and just talk. How we built a real-time, interruptible voice assistant on the Gemini Live API's native audio: streaming PCM both ways, barge-in, gapless playback, and live transcripts.

Read article →
Web Development
The Metro Vancouver F&B business tech stack: point of sale, ordering, back of house, guest and the website

The Metro Vancouver F&B Business Tech Stack

Five layers (point of sale, ordering, back of house, guest, website) and which of them a café, a dining room, a taproom or a caterer actually needs.

Read article →
Cloudflare infrastructure architecture diagram for a business website

What Broke After Launch

Contact form spam, a verification widget that failed silently, missing visitor data, injection vectors. The problems that only show up once a business site is live.

Read article →