Architecture, modernization and mentoring
Upgrade the system your business depends on without stopping operations: automated tests check every step before each version switch. For business and technology leaders.
Free first conversation · Reply within 1 business day
Regal Rexnord (NYSE: RRX):
PHP 5.6 → 8.4, 10 versions, zero-downtime cutovers
Two systems migrated from CodeIgniter 2 to 3, with unit-test suites
Quilt Software / PosNation:
Test suite from 540 to 1,455 in 3 weeks
Claude Code used to reverse-engineer undocumented legacy code
Regal Rexnord (NYSE: RRX): PHP 5.6 → 8.4, 10 versions, zero-downtime cutovers
Two systems migrated from CodeIgniter 2 to 3, with unit-test suites
Quilt Software / PosNation: test suite from 540 to 1,455 in 3 weeks
Claude Code used to reverse-engineer undocumented legacy code
The system runs on an unsupported PHP or framework version, and every server update turns into a risk.
There are no tests and no documentation: only one or two people understand the business rules, and they may leave.
A rewrite from scratch looks like the way out, but operations cannot stop or wait for the new version to catch up.
Every deploy is manual and tense, and the team spends more time firefighting than delivering the roadmap.
An inventory of versions, dependencies, integrations and untested areas, with a modernization plan in steps ordered by risk and by business value.
Unit and integration tests that pin down the current behavior of the system before any change, so every step is validated against what already works.
Version-by-version upgrades of PHP, CodeIgniter and Laravel, fixing the incompatibilities at each jump while the system stays in production the whole way.
Specs, business rules and OpenAPI contracts recovered from the code with AI and reviewed by people who know the operation, so the system stops depending on one or two people.
Deploys stop being manual: a pipeline with automated tests, static analysis and pre-commit hooks, and no change reaches production without passing the same criteria.
A Docker development environment that is the same for the whole team and a cutover runbook with acceptance and rollback criteria set before the date.
See also: AI coding agents, software architecture consulting, and custom software development.
Upgrade of critical applications from PHP 5.6 to 8.4 and migration of two systems from CodeIgniter 2 to 3.
Tests and quality gates on a legacy PHP (~160K lines) and React/TypeScript (~88K lines) platform shared by ~18 engineers.
Related experience: multi-tenant user management system built from scratch on Laravel 12 and Vue 3 for a retail-pricing platform in 20,000+ locations.
Related experience: the same method, Spec-Driven Development with AI coding agents, as the engineering operating model.
Facing a similar challenge? Tell us the problem and the timeline.
Free first conversation · Reply within 1 business day
Legacy modernization is upgrading an aging system the business uses every day, in small steps and without stopping operations. Spec-Driven Development is the method: before any change, current behavior is written down and covered by automated tests. Hire it when the version is out of support, every update is feared or the system has stopped evolving.
We read the code and map versions, dependencies, integrations and untested areas. You receive a risk report and the modernization plan in steps, with an estimate for each one.
We recover the business rules by reverse engineering and write the tests that pin down current behavior. You receive the specs, the test suite and a working CI.
We modernize one module end to end, from upgrade to cutover, to validate the method on your system. You receive the module in production and a revised estimate for the rest.
Version-by-version upgrades, with frequent releases and the system live. A step only moves forward with the tests passing. You receive each step in production, with its rollback runbook.
Documentation, pipeline and standards stay with your team, with mentoring to keep the pace. Evolution of the system can continue under a monthly retainer.
| Scope | Typical timeline | Engagement model |
|---|---|---|
| Legacy assessmentInventory of versions, dependencies and risks, a modernization plan in steps and an estimate | 1–2 weeks | Fixed-scope project |
| Modernization pilotOne module with tests, upgrade and cutover in production, to validate the method | 2–4 weeks | Fixed-scope project |
| Incremental modernizationLanguage and framework upgrades, tests, CI/CD and documentation, with the system live | 6–12 weeks | Fixed scope per step or time & materials with a cap |
| Ongoing evolutionFurther upgrades, continuous improvement and team mentoring | Ongoing | Monthly fractional retainer |
Timelines are typical ranges. Discovery defines the scope and the estimate of each project; the contract can be in USD or BRL.
Want to know which format fits your project?
Free first conversation · Reply within 1 business day
We use AI coding agents to understand and refactor legacy code faster, under one rule: no AI-generated change goes in without an approved spec, passing tests and human review.
Where AI comes inClaude Code reads undocumented code, maps dependencies and proposes specs and tests that describe current behavior.
Spec firstIn Spec-Driven Development, the expected behavior is written and approved before the agent changes a single line.
What humans reviewThe recovered business rules, the plan for each upgrade and every merge go through senior human review.
Quality gatesAutomated tests, static analysis and pre-commit hooks block any change that breaks current behavior.
Team adoptionPutting AI coding agents into your engineering routine is the subject of the AI software automation service.
Your dataClient data never trains models, and we work under NDA.
| Criterion | Rewrite from scratch | Incremental modernization | Keep as is |
|---|---|---|---|
| When it makes sense | The current architecture blocks the business or the technology has no upgrade path | The system works and holds business rules that must not be lost | The system will be retired soon or has low business impact |
| Operations during the work | Two systems in parallel until the switch | The system stays in production; each step goes live separately | No change |
| Return | Shows up only after the full switch | Shows up with every step delivered | None; maintenance cost tends to grow |
| Main risk | Undocumented business rules are left out of the new version | Requires test discipline and living with old code for a while | Unsupported versions accumulate security flaws and technical debt |
| Prerequisite | A complete specification of current behavior | Tests that pin down current behavior before each change | A retirement or replacement plan |
In most cases, modernize in steps: the system stays in production, the return shows up with every delivery and the accumulated business rules are not lost. A rewrite is justified when the current architecture blocks the business or the technology has no upgrade path. A 1–2 week discovery compares both paths using the data of your system.
With tests that pin down current behavior and upgrades one version at a time, each with a rollback runbook: that is how we took Regal Rexnord's applications from PHP 5.6 to 8.4 without downtime. The step-by-step is in our guide “PHP 5 to PHP 8 Migration Without Downtime”.
Typical timelines are: discovery in 1–2 weeks, a pilot on one module in 2–4 weeks and each modernization cycle in 6–12 weeks. Large systems are split into cycles, and every cycle ends with something in production. The biggest factors are the size of the codebase, the existing test coverage and the number of integrations.
That is the most common starting point, and the work begins right there: we recover the business rules by reverse engineering and write tests that pin down current behavior before changing anything. At Quilt Software / PosNation, we took the suite from 540 to 1,455 tests in 3 weeks, on a legacy PHP and React/TypeScript platform shared by about 18 engineers.
AI reads and proposes; a person decides, with senior review, and nothing reaches production without an approved spec and passing tests. AI coding agents such as Claude Code speed up reverse engineering, test writing and refactoring. At Regal Rexnord, we used Claude Code to understand undocumented legacy code. Client data never trains models, and we work under NDA.
Yes. Spec-Driven Development (SDD) means writing and approving the specification, with acceptance criteria, before the code. On legacy code, the first step is turning current behavior into specs and tests; from there, every upgrade has a clear target and an objective check. The full method is in our guide “What Is Spec-Driven Development (SDD)?”.
It is worth it when the system sustains operations and at least one of these signs shows up: a version without security support, risky deploys, difficulty hiring for the technology or new features that take longer every time. If the system has low impact or will be retired, keeping it as is can be the right call, and we tell you so in the first conversation.
The cost depends on the size of the codebase, the existing test coverage, the number of version jumps and the integrations. That is why we do not publish a price list. A 1–2 week discovery produces the plan in steps and an estimate for each one. The proposal can be fixed scope per step or time & materials with a cap, in USD or BRL.
AI-focused IT consultancy
Senior engineering from discovery to delivery: whoever designs the architecture also writes and reviews the code.
Master Timm · Updated on
In the first conversation we look at the versions, tests and risks of your system and tell you plainly whether the path is to modernize in steps, rewrite or keep it as is.
Free first conversation · Reply within 1 business day