Engagements

Work you can check.

Every engagement below is a real problem, a real approach, and a real outcome you can point to. Sectors are named; client names are on request and only shared with mutual introduction.

Payments · Series C

A release train that stopped derailing

Advise · Build · Run

41 min → 6 min Deploy lead time

The problem

Every deploy was a 45-minute negotiation. Rollbacks were manual, the pipeline had 2,300 lines of unreviewed Groovy, and the platform team had lost two principals in six months. Product releases slipped from weekly to fortnightly to "when it's ready."

The approach

Two-week assessment named the pipeline sprawl as the root cause, not the team. We rebuilt the delivery layer in eight weeks - GitHub Actions + Argo + a single golden path - and moved the existing 14 services onto it one at a time under live traffic. Three months later we handed the platform to their team; we stayed on-call for the following two quarters as they took over the pager.

The outcome

Deploy lead time from 41 minutes to 6. Change-fail rate from 22% to 4%. Two engineers hired back into the platform team on the strength of what shipped.

Health · Enterprise

One data platform instead of nine exports

Build · Run · Evolve

9 → 1 Systems of record

The problem

Analytics team was maintaining nine parallel exports from operational systems into a Frankenstein reporting warehouse. Every report was a spreadsheet lineage. Compliance was one audit away from a finding on data provenance.

The approach

We built one warehouse (Snowflake), one transformation layer (dbt), and one contract convention so operational teams could publish canonical data without asking the analytics team. Fourteen weeks from kickoff to first published contract; another eight to retire the last legacy export. Handover to their analysts happened alongside build - nothing landed after we left.

The outcome

Nine systems collapsed to one. Report authoring time cut by ~60%. Data provenance now traceable to source in the warehouse itself; audit finding closed.

Logistics · Scale-up

Cloud spend cut without a rewrite

Advise · Evolve

−38% Monthly cloud cost

The problem

AWS bill doubled year-over-year with no proportional growth. FinOps consultants had recommended a full replatform; the exec team wanted to avoid a nine-month diversion.

The approach

Four-week assessment identified 60% of overspend concentrated in five services - oversized RDS instances, orphaned load balancers, and a data pipeline running 24/7 for a job that fires twice a day. We right-sized without changing the architecture, added budget alarms tied to real cost owners, and IaC-ified the remaining click-ops so future drift is visible.

The outcome

Monthly bill down 38% in one quarter. No service outage during the migration. Cost budget adopted by their engineering managers as a standing item.

Fit

Who this is for.

We take on four to eight engagements a year. That constraint is intentional - we can only bring the depth if the room isn't crowded. If any two of the below feel like you, we should talk.

→ You need one team accountable across strategy, build and operations - not five vendors you personally coordinate.
→ You have a specific hard problem and you want the principal engineer who's solved it four times before, not their trainee.
→ You want the answer to include what will not work, in writing, before you spend.
→ You're willing to be uncomfortable when we tell you the risk before you asked.

Bring us the specific hard problem.

A short note about your system, your constraint, and what you've already tried. We'll reply within one working day whether we're the fit or not.

Book a call