Supply-chain optimization platform with an embedded AI assistant
As Engineering Manager/CTO I lead Riverbit's supply-chain optimization platform for pharmacy franchises: a modular .NET 8 monolith — 30+ domain modules, roughly 1,500 test cases — that imports transactional data from customers' upstream systems, computes forecasts and replenishment orders per SKU per warehouse, and serves a Vue 3 + TypeScript frontend.
The distinctive part: the platform embeds a Model Context Protocol (MCP) server, and the built-in LLM assistant (Claude) is a thin orchestration layer over that same authorized tool catalog — one authorization surface, enforced per tool call per warehouse, whether the caller is the in-app chat or an external MCP client.
Engineering discipline runs through the stack: per-customer client modules deployed through their own 8-stage pipelines, Playwright end-to-end tests that run against freshly built containers inside CI, OpenAPI-generated frontend types, and everything — app, database image, frontend, monitoring — shipping through the same tag-driven GitLab CI.
Project information
- Technology used: .NET 8, MCP, Claude, Vue 3 + TypeScript, PostgreSQL + TimescaleDB, Hangfire, GitLab CI/CD, Playwright
- Project goal: End-to-end supply-chain optimization with AI built in, not bolted on
- Date: 2024–2026
- Role: Engineering Manager / CTO