Value-stream automations backend
Python, PostgreSQL, JSONata, OpenAPI/Swagger, PyInstaller, React/TypeScript, Docker, Helm, Kubernetes, Jenkins, Artifactory, pytest, JWT/JWKS, RHEL 8/9, Rocky Linux 8/9
Resume
Led the port of a Clojure analytics backend to Python, replacing live third-party API reads with local PostgreSQL queries and reimplementing the automations engine — an HTTP request plus a JSONata transform, recorded as a measure against a metric — along with its encrypted secrets store and runtime credential injection.
Designed and shipped a secure cross-product launch flow — verification of a signed launch JWT against a remote JWKS endpoint, a durable session with a sliding activity-based TTL, and session transport hardened against XSS and CSRF, iterating between header- and cookie-based approaches (including CHIPS-partitioned __Host- cookies) based on threat analysis.
Packaged the product as a self-contained on-premises install — a PyInstaller build sourcing a hermetic interpreter, bundling frontend assets, API documentation UI, and all runtime resources. Built once natively on RHEL 8 and smoke-tested per target OS against a Helm-provisioned PostgreSQL, then published to an internal artifact repository. Validated across RHEL 8/9 and Rocky Linux 8/9.
Built the release-branch automation the team now uses to cut release and hotfix lines — an atomic script that cuts matching branches across the service and its deployment-chart repository, a versioning scheme so release builds never collide with mainline or overwrite the latest tag, a dedicated no-bump publish job, and a discoverable release runbook.
Hardened the service across its attack surface — externalized logging configuration and redacted a credential leak from log output, restored SSRF blocking in proxy-request handling, corrected multi-tenant secret scoping so private credentials resolve in previews without crossing tenant boundaries, fixed Content Security Policy to unblock inlined assets, patched a published dependency CVE, and hardened application settings against misspelled environment variables and cleartext secret exposure.
Enforced a product-wide naming boundary, preventing internal codenames and upstream vendor system names from reaching end users through UI text, API payloads, or run logs.
Improved CI throughput by parallelizing test and build stages, serializing mainline builds to make the version-bump push atomic rather than retry-prone, and routing dependency installation through an internal mirror.