Insights
Code, AI and culture
Lessons from building AI-powered software at speed.
Organising 50GB of files with vector embeddings — and an LLM safety net
How we brought order to a 50GB document archive for a leading European research university: semantic search with vector embeddings, four clear categories, a lifecycle folder for every project — and an LLM fallback for the files embeddings can't read.
Jun 1, 2026 · 5 min read
How we work: the AI-augmented SDLC at Xoredge
A small, senior team plus AI ships faster than a big one — if the engineering discipline holds. Here's our lifecycle, from discovery to support, and the best practices behind it.
Jun 1, 2026 · 2 min read
Building reliable AI agents with LangChain, LangGraph & AgentField
Agents are software, not magic prompts. Here's the engineering discipline — explicit state, evaluation, tracing and guardrails — we use to make them reliable enough for customers.
Jun 1, 2026 · 2 min read
Inside the Xoredge AI Platform: one gateway for every LLM
Why we built a single, self-hostable gateway in front of every model provider — and how it cuts cost, removes lock-in and makes AI reliable enough for production.
Jun 1, 2026 · 3 min read