A production & testing control system for green-hydrogen manufacturing
A green-hydrogen manufacturer builds high-pressure electrolysers from thousands of serialized parts, each routed through a mix of in-house and external-vendor steps, then physically tested on dedicated stands. Spreadsheets could not hold the truth of where every part was.
Custom build · low-code + custom engineering
The challenge
Manufacturing here is not a straight line. A single electrolyser is thousands of serialized parts, and each part follows its own route of operations, some in-house, many at outside vendors for steps like drilling, welding, and sintering. Finished parts then move to a completely separate world of physical testing on dedicated stands.
- Every part has its own route
Parts enter batches at different stages with different step counts. There is no single fixed route to follow.
- Work constantly leaves the building
Parts are dispatched to external vendors and received back, step by step. A spreadsheet can't track what's where.
- Testing is a separate spine
Physical tests on stands run downstream of manufacturing, and can only start once a part has finished its entire route.
How we built it
We built SpaceHydrogen: one production management system on a low-code app over a Postgres backend, with the heavy business logic in transactional database procedures. Everything reads and writes around a single spine, the execution matrix of one row per part per step.
- The execution spine: every part's route explodes into job steps. A part advances as its steps move from not-started to in-progress to complete. Its true status is emergent, never out of sync.
- Dispatch and receive: operators send parts to a vendor's next step and receive them back, generating packing slips and advancing each part automatically.
- A parallel testing spine: tests assign finished parts to physical stands. Readiness is derived from manufacturing, so a test only becomes runnable when its parts are done.
- A live scheduler: a test-stand timeline auto-prioritizes the queue, accounts for changeover and downtime, and recomputes when priorities change.
- Reporting on top: planned-work and throughput analytics read straight off the spine, so the numbers are always live.
See it in action
A walk through five core screens, reconstructed to mirror the live system. It advances on its own, or step through it yourself. Data shown is illustrative.
Work Orders
Active batches
18
Parts in flight
1,204
Due this week
6
| Order | Batch | Stage | Parts | Due | Status |
|---|---|---|---|---|---|
| WO-2041 | Stack assembly run | Assembly | 24 | Jun 18 | In progress |
| WO-2038 | Membrane coating | Coating | 60 | Jun 12 | In progress |
| WO-2035 | Bipolar plate run | Machining | 18 | Jun 09 | Blocked |
| WO-2030 | Cell sealing | Sealing | 42 | Jun 05 | In progress |
| WO-2026 | Endplate batch | QA | 12 | Jun 02 | Complete |
| WO-2019 | Gasket set | Cutting | 88 | May 28 | In progress |
Every manufacturing batch on one board
- A work order groups serialized parts heading through the next phase of production
- Live status on each: in progress, blocked, complete
- Part counts, planned dates, and notes at a glance, searchable and filterable
Results
Manufacturing and physical testing tracked together, end to end
Every serialized part's true state, emergent from its job steps
Dispatch, receive, and purchase orders across both, in one flow
A clean spine ready for genealogy, NCR, and the next phase of work
We build systems like this one: deep, real operations software that runs a business, designed around how the work flows, and yours to own and extend.
If your operation is this tangled, we can help
When a team is holding it together with spreadsheets and scattered tabs, we build the system that runs it.