Case study · Process design · Internal tooling · Documentation
Review Tracker
Construction-phase equipment reviews used to live in inboxes: untracked, unowned, unfindable. I built the tool that gave them a home — 2,344 reviews finalized, zero email threads — and the reporting center that proved the process worked.
- Work
- Equipment Spec-Review Tracker
- Role
- Process designer & builder (AI-assisted development)
- Stack
- Streamlit + SQLite · Excel export · internal deployment
- Features
- 3-step review wizard · sign-off locking · reporting center · bulk upload
- Scale
- 2,344 reviews finalized
During an equipment build-out, every spec deviation needs a reviewer, a decision, and a record. Email produced all three — but never in the same place, and never findable during an audit. The fix wasn’t heroic engineering; it was process design: a wizard that makes the right workflow the easy one, a lock that makes sign-offs mean something, and a reporting page that answers the manager questions before they’re asked.
Like Eyesight, it was AI-assisted development: I designed the schema, states, and rules; AI accelerated the code; I owned every behavior it shipped with. The recreation below mirrors the production layout — sidebar, KPI row, both tables, the works — with invented names and renumbered equipment.
04.1The tracker
Review Tracker
Active Reviews
| ID | Review type | Master code | Tool ID | Reviewer | Status | Issue date | Sign-off | Days in review | WW | Jira | Equip sections |
|---|---|---|---|---|---|---|---|---|---|---|---|
| ALPHA52_Mech_90 | Mechanical 90% | ALPHA52 | TOOL-A12 | C. Vega | not started | 2026-07-31 | — | 2 | 2026-WW31 | ✗ | 3 |
| CHARL02_Mech_90 | Mechanical 90% | CHARL02 | TOOL-C04 | M. Chen | not started | 2026-07-31 | — | 2 | 2026-WW31 | ✗ | 0 |
| SIERRA03_Mech_90 | Mechanical 90% | SIERRA03 | TOOL-S09 | A. Rivera | not started | 2026-07-31 | — | 2 | 2026-WW31 | ✗ | 0 |
| PAPA05_Elec_60 | Electrical 60% | PAPA05 | TOOL-P02 | S. Aldayem | not started | 2026-07-31 | — | 2 | 2026-WW31 | ✗ | 0 |
| DELTA23_Mech_90 | Mechanical 90% | DELTA23 | TOOL-D07 | D. Okafor | not started | 2026-07-31 | — | 2 | 2026-WW31 | ✗ | 0 |
| CHARL42_Elec_90 | Electrical 90% | CHARL42 | TOOL-C11 | A. Rivera | not started | 2026-08-01 | — | 1 | 2026-WW31 | ✗ | 0 |
| CHARL02_Elec_90 | Electrical 90% | CHARL02 | TOOL-C04 | D. Okafor | not started | 2026-08-01 | — | 1 | 2026-WW31 | ✗ | 0 |
| QUEBEC03_Mech_90 | Mechanical 90% | QUEBEC03 | TOOL-Q01 | D. Okafor | locked ✓ | 2026-08-01 | 2026-08-01 | 0 | 2026-WW31 | ✓ | 2 |
📦 Non-Masterfile Reviews
| ID | Review type | Master code | Tool ID | Reviewer | Status | Issue date | Sign-off | Days in review | WW | Jira | Equip sections |
|---|---|---|---|---|---|---|---|---|---|---|---|
| PAPA08_Elec_90 | Electrical 90% | — | TOOL-P08 | S. Aldayem | not started | 2026-08-01 | — | 1 | 2026-WW31 | ✗ | 0 |
Wizard: ① Intake — scope & documents → ② Technical review — findings & disposition → ③ Sign-off — locks the record. Locked reviews are read-only forever; corrections are new linked reviews, never edits.
04.2The reporting center
The report page existed for one reason: to end “how are the reviews going?” meetings. A command-center row, velocity charts global and filtered, and the load by equipment area — enough for anyone to self-serve the answer.
▦ Reporting Center
🚀 Project Command Center
📈 Project Velocity
📦 PCS Reviews by Unit Area
Summary Counts
| unit area | pcs count |
|---|---|
| Electrical | 232 |
| Gas & chemical | 214 |
| Vacuum & exhaust | 180 |
| Coating | 118 |
04.3From the user guide
The tool shipped with a 4-chapter guide — Starting a review, Completing a technical review, Sign-off and locking, Reports and exports — written for the busiest reader in the building. A representative passage, rewritten generically:
“Step 3 locks the review. Locking is not bureaucracy — it is the entire point. A locked review is the version of events everyone agreed to while they still remembered the events. If something changes later, open a new review and link it; never wish for an edit button.”
Documentation that people actually read is a support skill: five years of explaining hardware failures to non-engineers at the Genius Bar turns out to be excellent training for writing chapter two of an internal manual.