Financial charts and analytics displayed on a laptop screen

Home  /  Our Work

Case studies

Systems in production, with the numbers attached.

A selection of engagements from the last three years, described by industry — client names are withheld under our standard confidentiality terms.

Financial services

Replacing a loan origination system without a freeze on lending

A regional bank was running originations on a fifteen-year-old application that only two people understood. Three previous replacement attempts had stalled because each required a full cutover the business could not absorb.

We put an API gateway in front of the legacy system and migrated one product line at a time — consumer, then auto, then commercial — running old and new in parallel with automated reconciliation on every application. The last legacy route was retired in month eight.

8 moFull cutover, zero downtime
63%Faster application decisioning
0Reconciliation breaks at go-live
Financial dashboard with performance charts
Healthcare

A scheduling platform for 40 clinics and 1,200 providers

A multi-site healthcare provider had grown by acquisition and inherited four scheduling systems that did not talk to each other. Patients were routinely booked into the wrong facility, and the call center carried the cost.

We designed a single scheduling service with a shared provider directory and rules engine, integrated it with the three existing records systems, and rolled it out clinic by clinic over five months with a hypercare team on site for each wave.

1.2KProviders on one system
-38%Call center handle time
HIPAAAudited controls at launch
Clinician reviewing patient data on a screen
Energy

Cutting cloud spend 41% while doubling deploy frequency

An energy company’s platform had been lifted and shifted to the cloud in a hurry. Two years later the bill had tripled, deploys happened monthly, and no one could say which of the 190 running instances mattered.

We spent four weeks producing a full inventory and cost attribution, then six months rebuilding the platform as reusable infrastructure modules with autoscaling, a proper staging environment, and a deployment pipeline any engineer on the team could run. We trained their staff throughout and left.

-41%Monthly cloud spend
22xDeploy frequency
99.98%Uptime, trailing 12 months
Data center corridor with illuminated server racks
Logistics

Document extraction that removed 9,000 hours of manual keying

A logistics operator processed roughly 4,000 bills of lading a week by hand. An earlier attempt at automation had been abandoned because accuracy plateaued around 70% and nobody trusted the output.

We built an extraction pipeline that combines document layout parsing with a language model, and — the part that mattered — a confidence threshold that routes uncertain documents to a human review queue. Accuracy is measured continuously against a labeled holdout set.

96.4%Straight-through accuracy
9,000Annual hours returned
11 wkConcept to production
Operations team monitoring logistics data on large displays
In their words
“They told us in week two that half of what we had asked for wasn’t worth building. That conversation cost them revenue and saved us a year.”
Chief Product Officer — logistics client
Industries

Where we have the most reps.

We are generalists by training and specialists by accumulation. These are the domains where we arrive already speaking the language.

Financial services

Lending, payments, treasury operations, and the audit trails regulators expect to find.

Healthcare

Scheduling, care coordination, HL7 and FHIR integration, HIPAA-aligned architecture.

Energy & utilities

Field operations, SCADA-adjacent data pipelines, asset management, and reporting.

Logistics & supply chain

Dispatch, document automation, carrier integration, and warehouse systems.

Mountain range under a wide sky at dusk
Let’s build something

Tell us about the system that’s holding you back.

Every engagement starts with a free 45-minute working session. No pitch deck — we dig into your architecture, your roadmap, and your constraints, then tell you honestly whether we’re the right team for the job.