OBI — a St. Louis software studioOBIStart a project ↗

The big ones we have already built

12 engagements, from a pizza counter to a state government. Five written up in full below; the rest are listed with what they were. Where a client is under NDA, the sector is named and nothing else.

Food service · Web & mobile app development

Blue Box Pizza

A pizza counter taking orders from four sources at once — phone, walk-in counter and two delivery tablets — while the kitchen could only see tickets in the order they arrived.

  • Next.js
  • Stripe
  • Kitchen display
The problem
Tickets were sequenced by arrival time rather than by what the ovens could finish next. A twelve-minute bake could sit behind a two-minute item, pickup times were promised against a queue nobody could see, and the usual correction was to remake the order at the shop's cost. Adding another tablet would have added another unsynchronized source.
What we built
One order-ahead site, one consolidated ticket feed, and a kitchen display that sequences tickets by bake time and pickup promise rather than arrival. The key decision was to normalize every source into the same ticket format, so the kitchen no longer depends on which channel an order came from.
Where it landed
The queue on the display matches what the ovens can deliver. Quoted pickup times are reliable, remakes are down, and the display remains usable during Friday-night volume.
Construction · Custom software development

Galve Development Group

A construction group whose bids were built by copying and editing the previous project's spreadsheet, leaving no reliable record of how a signed proposal's numbers were derived.

  • Postgres
  • PDF pipeline
  • Change orders
The problem
Copying a prior spreadsheet is common practice, and it is how labor rates drift, markup becomes inconsistent, and change orders agreed on site fail to reach the estimate. The proposal the client signed had no traceable link to the estimate behind it.
What we built
Scoped line items with labor rates, cushion and markup as explicit fields; change orders that update the estimate; and a client proposal generated from the same data as the estimate. One data source produces all three documents.
Where it landed
A bid is completed in an afternoon rather than a week, and each signed proposal traces back through its change orders to the original estimate.
Millwork & doors · AI automation & agents

Galve Door Co.

A door installation company whose inbound texts and calls were answered personally by the owner, so leads that arrived while he was working went unanswered.

  • AI agents
  • Scheduling
  • CRM
The problem
Inbound demand came in by text, phone and web form throughout the day and outside business hours. Each contact needed a reply, a qualifying conversation and an appointment fitted to the installation schedule, and all of it depended on one person being available. Unanswered leads were lost.
What we built
An automated CRM in which an AI agent handles inbound text and voice conversations, qualifies the job, books an appointment against the installation schedule, sends confirmations and follows up, and hands the conversation to a person for decisions it should not make on its own.
Where it landed
Inbound leads are answered and scheduled without the owner's involvement, and his working day is spent on installations rather than on the phone.
Labs & instrumentation · Custom software development

X-Pro Labs

A laboratory in which samples were tracked on paper and instruments logged to the workstations they were connected to, so reconstructing how a result was produced was a manual exercise.

  • Python
  • FastAPI
  • MQTT
The problem
Chain of custody is a compliance record, and it existed on a clipboard. Instrument output stayed on the local machine, so linking a result to its run, instrument and operator required manual reconstruction — slow routinely, and unreliable under audit or dispute.
What we built
A sample pipeline that records custody at each handoff, instrument telemetry ingested over MQTT in real time, and results that carry their own audit trail — run, device and operator stored with the value.
Where it landed
Each result is traceable to its run, instrument and operator, and the trace is generated by the system rather than reconstructed afterward.
Insurance · AI automation & agents

Under NDA — Insurance

An insurer receiving claims by form, email and phone into separate systems, each with a different definition of a claim record.

  • Under NDA
  • Triage
  • Human in the loop
The problem
Time to first contact depended on the intake channel. Three intake paths meant three record definitions, three queues, and triage performed by whoever opened the item. Full automation was not acceptable because some decisions require a person.
What we built
A single intake and triage flow across all channels that scores and routes work to the appropriate desk, requires human approval above a defined threshold, and logs every decision. The boundary between automated and human decisions is explicit. Details withheld under NDA.
Where it landed
Shorter time to first contact, consistent handling regardless of channel, and an auditable decision log. Client name and figures withheld under NDA.
Also shipped

Seven more, in brief

Catering

Italian Hoagies

Large catering orders were the shop's highest-value business and its most error-prone: each was priced by hand over the phone and re-entered three times before delivery.

Plant-based kitchen

Vegan Pot

A plant-based kitchen maintaining the same dish in three places — the website, the allergen sheet and the weekly subscription plan — with the three copies frequently out of sync.

Real estate ops

Deloya Group

A real estate operator with ten years of property, vendor and draw history, none of it in a single system.

Benefits

Under NDA — Benefits

A benefits firm in which an advisor's readiness to sell had different answers depending on which system was consulted.

Residential

Under NDA — Residential

A residential operator managing a large portfolio property by property, with each site running leasing and resident requests its own way and no consolidated view.

Public sector · Mexico

Ressesa

Ressesa, a Mexican technology company, engaged OBI as its engineering partner on a large platform for the Mexican federal government.

Public sector

Under NDA — State Government

A Mexican state government whose services were nominally digital but still required residents to queue in person at offices that did not share records.

Your build would be number thirteen.

Start a build →