The solution
The engagement ran in two phases. AI-assisted development is what made both fit inside the time available.
Phase one, modernization and remediation
- Upgraded the customer portals from Angular 16 to Angular 21, a prerequisite for closing many findings rather than a separate project
- Migrated the Azure Functions layer from the .NET 6 in-process model to the .NET 10 isolated worker model
- Updated packages throughout and cut Angular bundle sizes by about 50 percent on average across the projects
- Remediated every vulnerability the client's commercial scanner (Snyk) reported, more than 620 in total
- Cross-checked against the open-source OSV database, which surfaced real vulnerabilities the commercial scanner had missed, and fixed those too
- Closed every outstanding penetration test finding
- Deleted Azure resources that were already dead, which alone took more than $25,000 a year off the bill
- Used AI to produce polished written summaries at each milestone, which the client's sponsor sent to leadership as they were
Phase two, one set of shared services (in testing)
- Converting the platform from five to six per-customer copies into a single multi-tenant set of services, so a new retail account means adding a database rather than a whole suite of Azure resources
- Retiring the duplicated Azure resources as the conversion lands, which is where the projected $78,000 a year comes from
Clearing one commercial scanner’s findings is the usual definition of done, so the OSV cross-check, which found real vulnerabilities the commercial scanner had missed, is what took the portals past a clean report.
The results
The scoped security work finished ahead of the client's plan, and the client credits the compressed timeline to the engineer's use of AI.
More than 620 known vulnerabilities (CVEs) eliminated, including findings the existing scanner had not reported
Every penetration test finding remediated
Portals on Angular 21; Azure Functions moved from .NET 6 in-process to .NET 10 isolated worker
Angular bundle sizes about 50 percent smaller on average
Over $25,000 a year in Azure spend already removed
Five to six copies of the codebase becoming one set of shared services, projected to save about $78,000 a year in total
The consolidation wasn’t in scope and only became possible because AI-assisted delivery finished the scoped work early. The conversion is in testing; the $78,000 stays a projection until it is in production and a full billing cycle has run.
“I attribute much [of] the quality of this work and especially the compressed timeline to Riel's mastery of AI, and his ability to apply AI to solve problems and get things done... It has been very refreshing to work with someone who 'gets it.'”







