Case study
AI-Powered Pipeline & Frontend Modernization
Feb 2026 – May 2026
Modernizing 50 ETL pipelines and 14 production apps with AI agents — and proving it was safe to ship.
Problem
A large investment services platform was running 50 ETL Python pipelines on Windows VMs with Apache Airflow, and 14 React applications on an aging stack — slow pipelines, manual remediation, high risk of regressions on every change.
Approach
- Used AI to explore and analyze the existing application and codebase before touching anything
- Designed and ran a remediation process, then progressively automated it — reaching 80% automated remediation while holding regressions to near zero
- Migrated pipelines from Windows VM/Airflow to an OpenShift (Kubernetes) environment
- Adopted the latest React features and React Compiler across all 14 applications
- Migrated to the latest NX monorepo version and implemented Module Federation for microfrontends
- Increased automated test coverage for key user journeys (Playwright)
My Role
Led the team through the remediation effort — set the process, validated AI-generated changes before they shipped, and was the point person ensuring output quality stayed production-grade throughout.
Outcome
- 80% of remediation work automated via AI-assisted workflow
- Pipeline execution time cut in half
- Near-zero regressions across 14 applications during the migration
- Stack modernized: Airflow/Windows VM → OpenShift; legacy React → React Compiler; NX + Module Federation