26 projects across AI agents, integration, analytics & enterprise apps
These are my most recent deployments — the last two years or so of building, integrating, and shipping. If you need older references or want to dig into a specific domain, just ask: there's plenty more in the logs.
What I enjoy most isn't maintaining the known — it's the unknown. The ambitious brief, the legacy system no one has touched in a decade, the integration that "can't be done". Those are the missions I'd volunteer for twice.
To boldly build what no one has built before. Or at least, what hasn't been built here yet. 🖖
33 active skills, 12 MCP integrations, 16 enforcement hooks, 7 ADRs, right-model routing (Scout / Builder / Advisor) and a full session lifecycle protocol.
→ Explore the frameworkThe watcher polls flow run history on a schedule, de-duplicates by run, ignores transient retries and alerts a phone only for new failures within working hours — a script rather than an agent, because nothing in it needs reasoning. The guard is a pre-tool hook backed by a regression suite built from real incidents and then re-tested under real session conditions, which exposed three gaps the lab cases had missed; its coverage limits are documented beside it.
Enables AI-assisted ABAP development — source navigation, debug sessions, transport management and RFC calls — entirely from the Claude Code CLI. Removes the context-switching between IDE and AI tools.
Agents can create, update and close tickets autonomously, classify incoming requests and draft responses — closing the loop between AI reasoning and real ITSM workflows without manual handoffs.
Reduces manual security triage time and improves detection accuracy over rule-based filters.
Delivers unified KPI reporting across ES, UK and US operations through Power BI — replacing manual exports and disconnected local reports.
Eliminates manual document search for operational teams — answers that previously took hours now return in seconds, with source citations and compliance traceability.
Enabled self-service reporting and ensured high availability. Historical tables as backups reduce reprocessing time and costs.
Achieves +90% accuracy, user-friendly UX/UI, restricted access, scalability, multilanguage support, and source document citations.
Microsoft Forms triggers a Power Automate flow calling a SAP Integration Suite iFlow — bridging cloud and on-premise via RFC and Cloud Connector to orchestrate resets through ManageEngine Plus API.
Enables product simulation, stock checks, pricing, and discount application via enterprise system integration. Azure Functions improve performance.
Centralises documents, ensures segregation of duties, and streamlines approvals with clear status tracking.
Demonstrates hybrid integration combining modern API design with legacy system interoperability.
Handles first-level support queries, ticket creation, and knowledge base lookups using RAG over internal documentation. Reduces ticket volume and accelerates resolution times.
Integrates with SAP BTP via SAP SDK, enabling offline data capture and real-time synchronisation without MDM infrastructure. Production build signed and validated end to end on an Android emulator against the production identity provider (password grant via ROPC), including the failed-login path.
Canvas app with multi-step approval flows, notifications through Teams and email, and SAP-replicated cost-object master data cleaned by real activity rather than formal status. Alongside it, an executive reporting service renders branded PDF summaries and sends them from a mailbox-scoped application identity (Exchange RBAC for Applications) — so the app can mail as one address only, not as anyone in the tenant. Bilingual admin and user manuals ship with it.
Inventory of every extractor against the metadata catalogue, risk assessment of the restricted ODP-RFC path, diagnosis of timing-out sales extractors and a vendor-sanctioned replication target evaluated end to end. In progress; the point is a dated, evidence-backed migration plan rather than a last-minute rewrite.
Built to automate manual research and apply data science to real financial data — from ingestion and modelling to interactive visualisation, with public deployment.