Engagements

Who was stuck, and what changed

Three engagements, written the way we scope them: the people affected, the work, the result, and the tradeoff we named before starting. Client names are withheld where the contract requires it.

$3.1M+
Client revenue contributed across engagements
400+
Automated tests running nightly in production pipelines
500+
Organizations served by search infrastructure we built
2024
Innovation award in a regulated utility environment
01
Regulated utility · Data platform
Data engineeringRow-level securitySQL Server

Four business units were waiting on numbers they could not trust

Legacy reporting broke every time the source schema changed. Analysts in four business units rebuilt the same extracts by hand, and the finance close depended on figures nobody would sign off on without checking them twice.

What we did

  • Migrated the legacy operational data store to an analytical data store, modeled for the reporting the business actually runs.
  • Built an abstraction layer between reporting and the source schema, so a schema change no longer breaks a dashboard.
  • Designed the row-level security model against Active Directory groups the client already maintained, with schema-per-group tenancy.
  • Automated the support workflow around data corrections and handed the runbook to the internal team.
60%
Faster support resolution
4 units
Reporting on one source of truth
2024
Formal innovation award
Tradeoff we named in week one

The abstraction layer adds a migration step the reporting team has to plan for. We scoped it, dated it, and ran it with them rather than leaving it in a footnote.

02
Healthcare SaaS · QA and search
SerenityJSJenkinsBedrockOpenSearch

The inherited test framework was unsalvageable, and the release date held

A platform serving more than 500 organizations had a release on the calendar and a vendor-built test suite that failed for reasons nobody could reproduce. The QA lead was spending her week triaging false failures instead of testing the new work.

What we did

  • Rebuilt the framework on SerenityJS using the screenplay pattern, starting from a four-month proof of concept the client could review before committing.
  • Wired 400+ end-to-end and API tests into Jenkins for nightly execution, with automated error alerting and dashboards a product owner can read.
  • Designed and deployed AI-powered search on AWS Lambda, Bedrock and OpenSearch Serverless, defined in CloudFormation and released through the same pipeline.
4 mo
Proof of concept to production
400+
Automated tests running nightly
$1.8M
Contract extension secured
Tradeoff we named in week one

Nightly runs need a stable test environment. That was a dependency on their platform team, and we said so before the first sprint rather than reporting it as a blocker later.

03
Enterprise platform · Onboarding automation
Process automationCloud onboardingSQL optimization

Setting up a new customer took a week of manual work

Every new customer environment was assembled by hand from a checklist. The implementation team absorbed the delay, sales absorbed the wait, and a single missed step meant reworking the tenant after go-live.

What we did

  • Mapped the onboarding checklist to the systems it touched, then automated the provisioning path end to end.
  • Optimized the queries behind the slowest steps and removed the manual data loads entirely.
  • Documented the new path and trained the implementation team to run and extend it without us.
90%
Less customer setup time
0
Manual data loads remaining
Owned
Handed to the internal team at close
Tradeoff we named in week one

Automating the checklist froze the process. Any future change to onboarding now needs a code change, which is a deliberate trade for consistency.

Contact

Send the constraint you are stuck on

A founder reads every message. We reply within two business days, and we will tell you if this is not work we should take.

Start a conversation See capabilities