Systems that had to work
on Monday morning.
Every project below is software that runs a real operation — not a prototype and not a portfolio piece. Client names and operational details are withheld under confidentiality. The engineering decisions are described exactly as they were made.
Integration & data movement
Getting systems that were never meant to talk to each other to agree, on a schedule, without a person in the middle.
Knowing where the freight actually is
Inbound shipments were tracked by opening carrier websites one tab at a time. We replaced the tabs with live carrier integrations and a workflow that routes exceptions to the person who can fix them.
Automating a system that has no API
The warehouse platform offered no write interface, so every transaction went in by hand. We automated the screens themselves — carefully enough to trust with live inventory.
Turning a flat order backlog into a plan the floor can pick
A daily backlog feed said what had been ordered — not what could be built, released or picked. We built the pipeline that turns it into a wave plan, a queued pick order, and a screen that says where each one got to.
Variances that only close when they're actually gone
Two systems held different quantities for the same parts, and the disagreement arrived as a spreadsheet attached to a recurring email. We turned it into a register that remembers — and that nobody is allowed to close by hand.
Return paperwork the receiving system accepts first time
Returns had to be announced to the warehouse system by hand-filling a hundred-column vendor template. We build the document from the check-in itself, check every line against the customer's item master, and track each upload through to accepted or failed.
Compliance, safety & controls
Software that enforces a rule rather than recording that somebody meant to follow it.
Audits that can't be quietly passed
Customer-facing quality audits ran on paper and a survey tool. We built a configurable audit platform with scheduling, escalation, approvals — and a gate that blocks sign-off when the scan doesn't match.
Equipment that locks itself out when it fails inspection
Powered industrial trucks were signed out on a clipboard, and the fall-protection harness that some of them require was a separate honour system. We built a badge kiosk where the truck you pick decides the inspection you must pass — and which refuses to release anything that fails.
Financial controls on material leaving the building
End-of-life material left the site on paperwork typed by hand. We rebuilt the chain end to end and added the control nobody had: detection of inventory added after the paperwork was already filed.
Adding a permission shouldn't mean altering the user table
Access was a boolean column on the user record, so every new screen was a schema change. We replaced it with roles that map to screens, credential checks that verify a second person without elevating the first, and a suspension path that fails closed.
Operations platforms
Whole systems that replaced spreadsheets, inboxes and the knowledge that lived in one person's head.
One application that replaced a decade of spreadsheets
A warehouse ran its floor on shared spreadsheets, emailed reports, and tribal knowledge. We replaced all of it with a single internal platform — and kept it running while people worked.
Exceptions that open their own case file
A damage or count discrepancy is a claim, and a claim is worth whatever evidence is attached to it. We built the casework — photo capture at the point of discovery, a case that assigns and escalates itself, and a PDF report that goes out with the pictures already in it.
A request queue where the approvals set the status
Internal software requests arrived by email and were approved in conversation. We built an intake form, a board and a three-role approval gate where the votes move the card — and gave requesters a private link so they can follow and edit their own request without an account.
Knowing you'll be out of stock before you are
Packaging and floor consumables were tracked on a shared spreadsheet, and the first reliable sign of a shortage was an empty shelf. We built a scanned stockroom with a background service that projects each item's stockout date against its lead time — and emails only when something new goes wrong.
Hardware & the physical floor
Printers, badge readers and the awkward gap between a web browser and a device on a bench.
Making eight label printers stop locking up
Label printers wedged under load and needed power-cycling. The fix was understanding what the hardware actually does — and sending it 500 times less data.
Making a badge reader work in a browser that can't see it
Browsers are not allowed to talk to smartcard hardware, and the kiosk desktops were centrally managed. We built a small local service that bridges the reader to the page — then packaged it so it installs, restarts and heals itself across a fleet.
Running it in production
The unglamorous half: knowing it's broken before a user does, and shipping to something nobody can stop using.
Knowing the platform is slow before a person walks over to say so
An internal platform with almost two hundred routes had no way of reporting on itself. We instrumented it from the inside — request timing, worker-thread introspection, user analytics and a live console viewer — without letting the monitoring break the thing it monitors.
Deploying to a system that nobody can stop using
A platform the whole site works on can't have a maintenance window. We built a two-slot deployment pipeline that refuses to stop the running version until the new one is proven up — plus an in-app changelog and a request board so users can see what changed and what's coming.
Why these read the way they do
Case studies usually lead with a number. These lead with the problem, because the numbers belong to the clients who paid for the work.
What we can't tell you
Client names, site locations, system names, volumes and dollar figures are all covered by confidentiality obligations we intend to keep. If a competitor's case study names a client, either they got written permission or they shouldn't have.
We'd rather show you the reasoning than borrow credibility from a logo. On a call we can go considerably deeper on approach, architecture and what we'd do differently — without crossing that line.
The same rule governs the numbers. Every figure on these pages is a property of the software itself — route counts, retry limits, poll intervals — because those are ours to publish. What the software did to a client's cost or headcount is theirs, and it stays with them.
Production, not proof-of-concept
Everything here ran daily, under load, with people depending on it during their shift.
The failure modes are the story
Anyone can describe the happy path. What matters is what the system does at 3am when a dependency is down.
No rip-and-replace
Almost none of this involved replacing a system. It involved making the systems already in place work together.
Recognize your operation in any of these?
A free assessment gives you a clear picture and a prioritized plan — no commitment.