blog.fichtmueller.org
// Production AI
Ollama Local LLM Mac Studio Production AI Self-Hosted Privacy llama qwen Apple Silicon AI Infrastructure

Ollama in Production: Running 70B Locally

Mac Studio M4 Pro with 48GB unified memory runs llama3.3:70b for reasoning tasks. Real latency numbers, model selection logic, and where local inference actually beats cloud.

Rene Fichtmueller / 2026-05-22 / ~2 min read min read
LLM Gateway AI Circuit Breaker Failover Production AI TypeScript Fastify Confidence Scoring Ollama Architecture

LLM Gateway Patterns: What We Learned After 50,000 Requests

Circuit breakers, confidence scoring, failover chains — an LLM gateway isn't a proxy. After 50,000 production requests through our internal gateway, here's what the patterns actually look like.

Rene Fichtmueller / 2026-05-21 / ~2 min read min read
RAG LLM Production AI Embeddings Vector Database Qdrant AI Engineering Chunking Retrieval OpenAI

Why Your RAG Pipeline Fails in Production

Tutorials show RAG working at 95% recall on toy datasets. Production systems hit 60%. The gap isn't the technology — it's three implementation choices that look harmless and aren't.

Rene Fichtmueller / 2026-05-20 / ~2 min read min read