Engineering notes on AI
Notes we are writing on the practice of building AI systems — implementation checklists, RAG architecture, agent safety, red teaming, eval-first engineering, and prompt injection defense. Each piece is grounded in the work, not in marketing.
AI Implementation
The AI Implementation Checklist
A practical checklist for moving an AI capability from prototype to production — covering data, retrieval, evaluation, guardrails, and operations.
RAG
Designing a RAG System That Survives Production
How to design retrieval, chunking, reranking, and evaluation for a RAG system that holds up under real usage — with grounding and access control.
AI Security
A Practical AI Red Team Methodology
How we scope and run red team engagements against AI systems — threat modeling, attack trees, and reproducible findings.
AI Agents
Agent Safety Patterns: Scopes, Gates, and Traces
Engineering patterns for building agents that can be trusted in production — permission scopes, human approval gates, and execution traces.
AI Engineering
Eval-First AI Engineering
Why evaluation should drive every decision in an AI system — and how to build an eval harness that actually catches regressions.
AI Security
Defending Against Prompt Injection
A practical guide to prompt injection vectors and the engineering defenses that actually reduce risk in production systems.
Common questions
The questions we get most often. If yours is not here, send us a note.
General
Services
Engagement
Security
Have a question we did not answer?
Send us your question. We answer honestly, including telling you when we are not the right partner for what you need.