The problem
Machines on a production line emit a steady stream of operational data. Without somewhere for it to land and something to read it with, that data is effectively invisible — problems are noticed when output is already wrong, and questions about what happened last Tuesday have no answer.
What I built
A complete dashboard and management application, owned end to end: ingestion from the machines, aggregation and storage, and the frontend the floor actually uses.
- Capture — ingesting machine data across the production line into a consistent schema, so heterogeneous sources become queryable in one place.
- Aggregate — rolling raw events into the summaries people ask for, rather than making every question a full scan.
- Analyse — a management and dashboard interface for monitoring current state and interrogating history.
- Forecast — projections built on the accumulated history, so the platform answers what is likely to happen next rather than only what already has.
Forecasting is what turns a monitoring tool into a planning one. A dashboard showing current state tells the floor that something is wrong now; a projection over the same data supports decisions before that point — anticipating demand and load instead of reacting once output is already affected. It also makes the historical data pay for itself twice, since the same series that powers the retrospective view drives the forward one.
AI-driven quality control
Alongside the platform, I contributed to the design of AI-driven quality control systems — including a feasibility study on automated defect detection in licence plate manufacturing: whether the defects that matter are reliably visible to a camera-based system, what imaging conditions it would require, and where the honest limits sit.
Most of the value of a feasibility study is in ruling things out early. Saying clearly what a vision system will not catch is more useful than a demo that works on the easy defects.
Some specifics are omitted — this is commercial work for a client's production environment.