AI Agents in Production
Why RAG fails in enterprise, and what to design instead.
Most enterprise RAG systems ship with broken metadata, no permission model, and brittle eval. We’ve rebuilt enough of them to know what actually works.
Read EssayEditorial · Engineering Essays
Field notes from the
forward-deployed engineering reality.
Technical essays, system teardowns, and field notes from production deployments. No thought leadership theater, just the patterns, failures, and decisions that survived contact with real operations.
AI Agents in Production
Evaluating LLM outputs in production: a practical framework.
Golden datasets, regression suites, drift detection, and the case for treating evals like CI tests.
Compliance Automation
How to build audit-ready AI systems for regulated workflows.
Append-only logs, deterministic orchestration, evidence chains, and the contract you sign with your auditor before you ship.
Enterprise Architecture
Integrating AI agents with legacy ERP systems.
Most ERPs were not designed for agentic access. Here’s the integration pattern we use to bridge them without breaking the system of record.
Compliance Automation
Reducing hallucination risk in document intelligence systems.
Hybrid retrieval, structural validators, refusal patterns, and why the prompt is the wrong place to fix grounding.
AI Agents in Production
Designing human-in-the-loop approval systems.
What reviewers actually need to see, how to close the loop into evaluation, and the UI patterns that make approval feel native.
Public Sector Digital Transformation
Notes from building a sovereign cloud deployment.
Key escrow, sealed enclaves, on-prem replication, and the constraints that don’t show up in any cloud vendor’s marketing.
Data Infrastructure
Lakehouse, warehouse, or both? A pragmatic decision guide.
When you actually need Iceberg vs. Snowflake vs. a hybrid, based on workload shapes we see in the field.
Digital Experience Engineering
Performance-first WebGL: shipping cinematic without ruining INP.
How to get SpaceX-grade hero scenes onto enterprise sites without blowing your Core Web Vitals budget.
Field Notes
What we learned from 90 days embedded in a ministry.
Forward-deployed engineering at the ministry scale: stakeholder mapping, political constraints, and shipping inside government time.
AI Systems Readiness Audit
Bring us your
most complex workflow.
In 7–10 working days, Verne maps your workflows, data sources, repetitive decisions, automation opportunities, and AI risk areas. You receive a prioritized roadmap showing what to automate, integrate, avoid, and build first.
Tell us what is broken. We will map the system.