AI Security • AI Infrastructure • AI Governance
Designing deterministic, production-ready AI systems with strong identity, containment, cost, and policy controls.
Positioning
AI control planes and governed RAG release pipelines
Cloud and Kubernetes platform architecture across AWS and GCP
Identity-first security, policy gates, and observability for live systems
US citizen
Why This Site
Derrick Weil is positioned here from source-backed material as an engineering-led architect focused on the Day-3 realities of AI systems: identity enforcement, isolation boundaries, cost governance, deterministic execution, and operator recovery.
LinkedIn Profile export
"I build production-grade cloud and AI infrastructure designed to survive real-world scale, cost volatility, and security threats."
Production-Grade AI Systems
"The goal of this book is not to make AI systems smarter. It is to make them explainable, controllable, and survivable when they are wrong."
Flagship Work
The homepage only features projects with strong alignment to governed AI systems, secure automation, platform controls, and production operations.
AI Governance
A modular ingestion control plane that treats RAG ingestion as software delivery instead of a single script.
This is the clearest direct evidence of AI governance positioning: policy gates, release controls, observability, and auditability around retrieval infrastructure.
See project detailAI Security
A serverless security pipeline for log ingestion, threat detection, enrichment, and escalation using AWS-native services and secure automation patterns.
It reinforces secure automation, event-driven controls, and practical security operations rather than abstract AI claims.
See project detailAI Infrastructure
A production-style Kubernetes lab that packages RBAC, multi-namespace design, autoscaling, CI/CD, and observability into a single platform exercise.
It is strong evidence for platform engineering depth that underpins secure AI infrastructure and governed runtime environments.
See project detailCloud Infrastructure
A Terraform-based AWS to GCP HA VPN with BGP for dynamic routing and failover.
This supports the cloud and infrastructure side of Derrick's positioning, especially where governed AI systems depend on cross-cloud platform design.
See project detailWriting / Book
The writing material is anchored in the Production-Grade AI Systems manuscript rather than generic AI commentary.
The manuscript defines production-grade AI around how systems behave under stress, degraded dependencies, retries, and operator intervention.
Production is a behavioral state: how a system responds when conditions are no longer ideal, assumptions are violated, and humans are forced to intervene.
The book argues that most systems reach prototype and even early operations, then fail when retries, ambiguous correctness, cost, and human recovery enter the picture.
Day-3 engineering asks: Can the system survive real usage without degrading into risk?
The core control catalog emphasizes identity, sandboxing, output handling, retrieval integrity, bounded retries, cost budgets, and operator runbooks as non-negotiable constraints.
The controls define outcomes, not tools.
Next Step
The site is structured for recruiters, engineering leaders, and consulting clients who need evidence of controls-oriented infrastructure depth.