Shared MCP gateway and AI agent service for the department's Atlassian toolchain
GenAI Systems & AI Platform Engineering
- Architecting a custom MCP server as a shared gateway to the Atlassian toolchain for 15–20 engineers across the department, plus the AI agent service sitting in front of it — owning the security design for both: JWT auth, prompt injection prevention, access control
- Designed a GenAI workflow that turns plain English into ready-to-run test scripts in the team's internal DSL, grounded in ETSI and 3GPP specs, reducing script authoring time from weeks to hours
- Built the retrieval system behind it to run fully offline on Windows, indexing thousands of pages of specification documents — PDF ingestion, regex-based chunking, three parallel vector stores on LlamaIndex and LanceDB, shipped as a standalone executable through Nexus
- Built an internal AI software delivery system that runs specialist agents across the SDLC, from analysis and scoping through to execution, verification, and release, with a separate debugging subsystem for structured fault diagnosis
- MCP
- LlamaIndex
- LanceDB
- RAG
- Prompt engineering
- JWT
- Python
Jenkins estate migration and automated test reporting
Platform & CI/CD Infrastructure
- Migrated the department's Jenkins estate to new infrastructure — 350+ pipelines, 150+ build nodes, terabytes of build history — cut over in a planned maintenance window
- Designed the target architecture and stood up the servers: Docker containerisation, nginx reverse proxy, PKI for secure access
- Replaced a manual test reporting process with an automated Jenkins-to-Jira Xray pipeline now carrying hundreds of test executions across 10+ test plans, including a Python layer that parses proprietary XML and zipped archives into JUnit XML
- Wrote the declarative Groovy pipeline tying it together, from artifact retrieval through to publishing in Xray
- Jenkins
- Docker
- nginx
- Groovy
- Python
- PKI