Case studies
Twenty years of engagements sit behind this page — most under someone else’s logo and someone else’s NDA. The record is below. The system I can open all the way up is my own: the agent fleet that runs this shop daily.
The record
Figures out of other people’s production are the easiest thing in this trade to inflate and the hardest for you to check. The numbers stay with the clients who own them; what is listed here is the work.
When I say a system will hold, it holds. What I will not do is hand you a figure out of someone else’s production and ask you to take it on faith.
Track record
| Engagement | Domain | Measure | Result |
|---|---|---|---|
| Multi-region AWS migration (PCI) | Cloud | Availability | Multi-region cutover · US-East / US-West |
| Same migration | Cloud | Infrastructure cost | Consolidated footprint, lower run-rate |
| Same migration | Compliance | SOC II support | Real-time visibility + audit trails |
| Digital transformation (The Weiss Group) | Web | Delivery | CloudFront CDN + asset pipeline |
| Same transformation | Engineering | Pipelines | Automated testing, code review, deploys |
| Email infrastructure rebuild | Deliverability | Deliverability | Self-hosted Mautic · automated IP warming |
| Pharmaceutical B2B platform (RX Automation) | Compliance | Controlled-substance orders | Full audit trails (HIPAA-relevant) |
| NBA streaming app (Roku) | Media | Performance | Production performance monitoring |
| Retention analytics platform | Data | Visibility | Looker / Tableau / Grafana + automated alerts |
| Deployment architecture | Ops | Release safety | Blue-green deploys with DNS failover |
Every row is a real engagement. References and fuller write-ups available on request — [email protected].
Selected stories
Cloud · PCI
As Senior Software Engineer at PCI, I moved production to multi-region AWS across US-East and US-West — a cutover with no downtime budget to spend, real-time visibility and audit trails built in for SOC II, and roughly 40% off the infrastructure bill.
Web · The Weiss Group
As Application Development Manager, I led a digital transformation that rebuilt delivery on a CloudFront CDN — and left behind automated testing, code review, and deploy pipelines so it stayed fast.
Email · deliverability
A deliverability crisis had put most of the list out of reach. Self-hosted Mautic with automated IP warming brought it back — on sending infrastructure the business owns rather than rents.
Live practice
Not client work — my own operation, run daily as an agent fleet. It is the standing test bed for everything on the services page.
Claude agent sessions across project folders — this site, AWS infrastructure, campaign tooling, quant research — each registered in a session registry, opened and closed by playbook-driven rituals.
A mem0 memory service every agent reads and writes — on the order of 150 durable facts per project namespace — so session 21 starts where session 20 stopped.
Per-project dispatch channels: agents leave work for each other in mailbox files, and a hub session routes anything that crosses project lines.
Every durable write goes through tooling that rewrites, fsyncs, and re-reads the file byte-for-byte before it counts. Agents that write files need the same guardrails as agents that move money.
Own-ops practice, labeled as such — shown here because it is the one agentic system I can demo end to end without an NDA.
Methodology
Monitoring and analytics go in with the first deploy. If it isn’t measured, it didn’t happen — and it can’t be defended later.
Availability, latency, cost, deliverability. Quality is monitored through architecture, testing, and analytics — and forecasts get graded on calibration, not on how confident they sounded.
Client figures stay with the clients. The one system on this page I can open end to end — append-only logs, failure-mode catalog, invariant registry — is my own fleet, and it is labeled as own practice, not client work.
Tell me what you’re trying to do. First conversation is free.