KEELEN CARRERA

Restaurant managers juggle inventory, ordering, and stock tracking across disconnected tools. Dine Flow's goal was one platform covering all three.

The Express/TypeScript/Prisma stock router, end-to-end: full CRUD on inventory records, expiration-date calculation from shelf life, and conditional logic branching between local stock creation and an automated supplier reorder request. Inventory decrements automatically the moment an order is placed. Schema design and the frontend were team-built; the stock router was mine.

The GET-by-ID endpoint was returning every stock item instead of the one requested — traced to referencing the wrong ID field from the database, fixed by mapping the correct one.

Deployed to Netlify and demoed live to the cohort in April 2025. This same codebase is what Case Study 01's security review is built on.

CRM and ERP org records drift out of sync over time — mismatched, inactive, or orphaned records that break clean bidirectional sync between systems.

A reconciliation engine in MuleSoft/DataWeave classifying every record as matched, inactive, orphaned, or missing a code, behind a real HTTP listener secured with bearer-token authentication I configured explicitly as a policy — not an Anypoint default.

A mapping error silently connected sales-rep category CUS to members from category ACT — no error thrown, just members quietly assigned to the wrong rep. Reworked the mapping conditions to properly separate all three categories, then re-tested against sample data to confirm.

This is the one workflow in the 10-pattern repo that was actually run and tested — via Postman against the live endpoint, checking that response data matched the source, not just a 200 status. The other 9 patterns are documented architecture and DataWeave designs using fictional data.

Detection logic doesn't click from reading alone. Before I touch a live SIEM, I wanted to build the habit of reading, evaluating, and mapping real detection rules against the adversary behavior they're meant to catch.

Researched and curated 5 Sigma detection rules mapped to MITRE ATT&CK techniques — including a brute-force SSH rule targeting T1110.001 — reviewed each rule's logic line by line rather than treating them as a copy-paste library, and documented adversary context and hardening guidance for each one.

These rules are curated and studied, not authored from scratch — I selected, reviewed, and learned from existing detection logic rather than originating it. Wazuh deployment and Atomic Red Team validation against a live attack/target VM pair are the active next phase, not yet started.