Chantilly, Virginia.
I build tests that can actually fail.
I've spent ten years on federal and enterprise systems that can't quietly break: health programs, regulatory reporting, data platforms, and now public-facing AI assistants.
The problem I keep finding is the same everywhere. Suites that pass without proving anything. A test that can never go red isn't a test. It's a receipt.
Vision Vault Systems is how I fix that for other teams. I do the work myself, with a small senior team, and I leave once your people own it.
Ferkat
Founder and principal consultant
| Before | After | What changed |
|---|---|---|
| 30% | 97% | Regression pass rate on a federal suite, after I rebuilt the framework. |
| 5 days | 6 hrs | ETL validation cycle on a federal data program. |
| 136 | 2 | Hard-coded waits removed from a federal healthcare regression. |
| none | 150+ | Databricks tables under validation, from a framework built from zero requirements. |
The pass rate went down first. That was the correct direction.
On a recent enterprise suite I checked every assertion for one thing: could it ever fail? I deleted the ones that couldn't and replaced them with checks grounded in real data. Defect detection went up.
I use AI every day for writing, migrating and triaging tests. I never let it decide what counts as verified. Every framework I build, and QAVision itself, is designed around that line.
Where I've done it.
Client names are withheld under engagement agreements. Read the case studies → · Full resume →
- 2025 to now
Test automation and quality architecture, federal programs
AI assistant validation, a Databricks validation framework built from zero requirements, and Salesforce Lightning frameworks with client teams trained to extend them.
- 2024 to 2025
Senior SDET, federal enterprise program
Rebuilt a Selenium and Cucumber framework, added GitLab CI with Docker and Allure, led JMeter performance work and Section 508 testing.
- 2016 to 2024
QA manager, automation lead and performance architect, federal health
Full-stack automation across regulatory workflows, load modeling for enrollment surges, and IV&V on Jenkins pipelines.