Observability for LLM apps: Logs, Traces, and Feedback
The observability signals that matter when LLM features fail silently, drift in quality, or become too expensive.
6 min
Writing · Page 4 of 4
Field notes on production AI, search, and shipping.
Operator notes with concrete tradeoffs: what I built, what broke, and what I’d do again. 40 posts and counting.
RAG on Azure with pgvector: Practical Patterns for Production
A practical architecture for building production RAG systems on Azure with .NET, Postgres, and pgvector.
7 min
.NET + Azure OpenAI: A Production-Grade Starter Kit
A production-minded starting point for wiring Azure OpenAI into .NET apps with the right operational foundations.
5 min
Semantic Kernel vs custom orchestration in .NET
A pragmatic comparison between using Semantic Kernel and building your own orchestration layer in .NET.
5 min
ML.NET vs Python in real teams
How to decide between ML.NET and Python based on team strengths, operational cost, and the type of ML work.
5 min
ONNX Runtime patterns: perf without pain
The ONNX Runtime usage patterns that matter most for startup cost, memory behavior, and inference speed.
5 min
Vector databases without the hype
Why most teams should start with Postgres and pgvector before reaching for a dedicated vector database.
5 min
Prompt security for enterprises
A practical guide to defending enterprise LLM systems against prompt injection, tool abuse, and secret leakage.
6 min
Benchmarks that matter: latency, throughput, dollars
Which benchmarks actually matter for AI products when user experience, throughput, and cost are on the line.
5 min
Legacy to AI in 90 days: the migration guide
A 90-day plan for moving from legacy systems to a focused AI win without betting the company on a transformation project.
6 min