Structure and traceability for a high-stakes medical release

FluidAI Medical develops data-driven surgical solutions for general and gastrointestinal surgery. When the software touches the operating room, the standard for quality isn’t “good enough” — it’s verifiable, traceable, and repeatable.
‍
With a multi-feature release on the horizon, FluidAI Medical needed a QA process that could match the complexity of their platform: ten features in development across separate branches, tested across multiple browsers and screen sizes, with every failure documented and every result traceable back to the source.

That’s where QA flow came in.

The challenge

FluidAI Medical’s release was not a single, monolithic deployment. It was a coordinated rollout of ten distinct features, each developed on its own branch and requiring isolated testing before being integrated into the broader release.
‍
The testing environment added further complexity. Every feature had to be verified across multiple browsers and screen sizes, ensuring consistent behavior regardless of how surgical teams accessed the platform in the field.

Testing scope at a glance
Ten features tested across separate development branches. Manual test execution across multiple browsers and screen sizes. Bug logging with full traceability to GitHub Pull Requests. Test run results updated directly on each PR for developer visibility. Iterative back-and-forth testing cycles until each feature passed.
‍

Without a structured process, a release of this scope becomes a liability. With the right tooling and methodology, it becomes a controlled, documented, and auditable quality gate.

Feature testing: AI-generated cases, human judgment

QA flow’s AI-powered test case generation was used to create test cases for each feature, structured around the specific requirements and behavior of that branch. This wasn’t automated testing for its own sake. It was a way to ensure coverage was systematic, not dependent on memory or individual interpretation.

Test case creation

Test cases for each of the ten features were generated using QA flow’s AI tooling, covering expected flows, edge cases, and failure states. The combination of AI-assisted generation and manual QA expertise ensured that nothing fell through the cracks, and that the cases reflected real-world usage, not just happy paths.

Test case management

Once generated, all test cases were imported into TestMo, the team’s test case management tool, creating a single source of truth across the release cycle. Every test case was tracked, every result was logged, and the entire QA effort was visible and auditable from one place.

Manual execution & bug logging

Test cases were executed manually by QA engineers. For every failed test case, a bug was logged with full detail. Test run results were updated directly on the corresponding GitHub Pull Request, keeping developers informed in real time and eliminating the gap between QA findings and development response.

Testing was iterative. Features were tested back and forth across browsers and screen sizes until each one met the standard required to proceed.

All ten features were tested with full coverage and documented results per branch. Every bug was logged and linked directly to its GitHub PR, zero gaps in visibility. Check out QA flow.

Three days to production confidence

With feature testing complete, the next challenge was end-to-end verification. A smoke test suite was built to cover all ten features together — not in isolation, but as an integrated system behaving the way it would in production.
‍
The smoke suite was prepared using a combination of QA flow’s AI capabilities and accumulated manual QA expertise from the feature testing phase. This meant the suite reflected not just the intended functionality, but the real failure modes and edge cases that had surfaced during earlier rounds of testing.

Smoke testing

Smoke testing ran over three consecutive days, covering all features end-to-end. Each day’s results were reviewed, any regressions were investigated, and the team moved to production only once the full suite had passed consistently.
‍
The result: a release deployed to production with documented evidence of stability, not just a belief that it was ready, but a verified record that it was.

The three-day smoke test run covered all ten features before production deployment. The release shipped with full traceability and verified test coverage.

“When your software supports decisions made in the operating room, the testing process has to be as rigorous as the product itself. Islands built a QA solution that transformed our testing cycle from weeks to hours.”

Abdallah El-Falou
CTO, FluidAI

Why this process works for high-stakes software

Medical software demands more from QA than most domains. The cost of a bug isn’t a frustrated user or a support ticket, it’s potentially a clinical decision made on bad information. That reality shapes everything about how QA flow approached this engagement.

Principle 1: Traceability
Every test, bug, and result was documented and linked, from QA flow through TestMo to the GitHub PRs where developers worked. Nothing tested in a silo.

Principle 2: AI + Human Expertise
AI accelerated test case creation and ensured systematic coverage. Execution, bug analysis, and smoke suite design were driven by experienced QA engineers. Both are required.

Principle 3: Iterative Rigour
Features weren't signed off after a single pass. Testing repeated across browsers, screen sizes, and edge cases until results were consistent.

What this delivered for FluidAI

A release is only as trustworthy as the process behind it. For FluidAI Medical, that meant arriving at production deployment not with fingers crossed, but with documented evidence of what was tested, how it performed, and what was fixed along the way.
‍
QA flow gave the team the tooling to move quickly without cutting corners — and the methodology to make that speed defensible.