Wilmington, North Carolina · Est. 2018

Every change ships behind a flag.

Ossify Digital Engineering builds and maintains software for companies that cannot afford a bad deploy. Nothing goes out without a feature flag, which means a rollback takes seconds rather than an emergency release.

Start a Project Our Services

deploy-policy.yml

# Every change ships behind a flag.
rollout:
  strategy: "flagged"
  default_state: "off"
  rollback: revert_flag()  # seconds, not a redeploy

on_incident:
  notify: ["client", "on_call"]
  postmortem: required
  shared_with_client: true

since 2018

7Years in Wilmington

flagged

100%Changes behind a flag

p50

41sMedian rollback time

headcount

12Engineers, no offshore

The Problem

Most rollbacks are a redeploy under pressure.

Something breaks in production at 4pm. The fix is to revert the merge, rebuild, run the pipeline, and deploy — twenty minutes on a good day, longer if the pipeline is contended or the migration was one-way. During all of it the thing is still broken.

Feature flags solve this and get skipped anyway, because adding a flag to a change is extra work for a change that is probably fine. The discipline erodes exactly where it matters: small, obviously-safe edits, which is where a surprising share of incidents come from.

Every change we ship is flagged and defaults to off. Turning something off is a config change, not a deploy, and our median rollback is 41 seconds. It costs us maybe 8% more engineering time per feature and we do not present that as free.

A rollback that requires a build is not a rollback. It is another deploy, made worse by panic.

engagement.spec

What we commit to.

The same table goes in every statement of work.

Field	Commitment
repository	Client-owned organisation, from the first commit
test_coverage	Reported per pull request, never asserted verbally
flag_default	Off. Every change ships behind a feature flag
rollback_time	Under 60 seconds via flag revert
postmortems	Written for every incident, shared with the client
on_call	Ours during the engagement, yours after handover
handover	Runbook, architecture notes, and a live walkthrough

Our Services

Six services.

Twelve engineers in Wilmington. Nothing offshored or subcontracted.

v1.x · greenfield

Product Engineering

New applications from architecture through launch, in your repository, with flags and coverage reporting from the first commit.

8–24 weeks typical

v2.x · brownfield

Legacy Modernisation

Incremental replacement of systems that still earn money. We do not propose rewrites that pause the business for a year.

3–12 months

infra

Platform & DevOps

CI, infrastructure as code, observability, and the flag system itself. Backups verified by restore, monthly.

4–10 weeks

api

Integrations & APIs

Third-party and internal integrations built to alert loudly on failure rather than to drop records quietly.

2–8 weeks

audit

Technical Audits

Codebase, security, and delivery review with written findings you keep, including when the finding is that the code is fine.

2–4 weeks, fixed fee

sla

Maintenance & On-Call

Patching, monitoring, and paid on-call, with a written postmortem for every incident shared with you.

monthly, 30 days notice

How We Work

You get the postmortem

Every incident gets a written postmortem and you receive it, including the ones caused by us. Most vendors handle incidents in a call and never write anything down, which means the same failure recurs and nobody can point at why.

Test coverage is reported per pull request, as a number in the PR rather than a claim in a status meeting. Where coverage drops we say so and explain the trade.

Handover is a runbook and a live walkthrough, not a zip file. On-call is ours during the engagement and yours after, and we train for that transition rather than assuming it.

Method

Four stages.

Scope

Architecture, risks, and a written estimate with the assumptions listed separately.

Build

Two-week increments, flagged and deployed continuously to an environment you can see.

Harden

Load testing, observability, and runbooks written before launch rather than after the first incident.

Operate

Paid on-call, monthly restore tests, and a postmortem for every incident.

Who We Serve

Companies where downtime costs money

Logistics, healthcare administration, financial services, and industrial software across the Carolinas and remotely across the US.

Engagements from about $45,000 to $900,000, and we will tell you when an off-the-shelf product covers the requirement, which happens more often than a software firm usually admits.

We decline roughly one enquiry in five. Usually a rewrite that should be an incremental migration, or a build that a configured SaaS product already does properly.

The Team