React Engineers Who Work Through the Backlog as Part of the Client's Team
We started in June 2025 with two React engineers. They got access to the client's VPN, GitHub and Jira, ran the project locally, and took their first tickets a week after the kick-off.
The client owned the backlog and assigned the tasks, and our engineers worked as part of its in-house team. They sent their changes to the client's code review, checked them on test orders and handed them to the client for acceptance before release. From the first month they also reviewed other developers' pull requests. Each engineer reported every day, ticket by ticket, so the client knew what every paid hour was spent on.
About two thirds of the tickets in the backlog were small changes for one state: add a field, add a check for an answer, stop showing a question. Each of them took a day or less. But a new field has to go through the questionnaire, the validation rules, the order API and the client's internal tools, otherwise the staff don't see it in the order and it doesn't get into the document, so one engineer took such a change through all of these systems, covered it with tests and released new pages and questions behind a feature flag.
About a third of the time went into 27 large tasks that took more than a week each. These were rules that had to work across all states or all products at once, new multi-page questionnaires for one state, and new flows. For example, the rule that a company cannot be its own member or officer is one check, but it had to work in the annual reports of every state and every type of company.
The engineers were hired as extra hands for the backlog, but they were also able to handle these complex tasks. They worked on a new way of building questionnaires, built the pages of a new questionnaire for converting a company to another entity type, and moved a shared page out of one questionnaire so that all formation questionnaires could use it. The largest state tasks were the address rules for all products in California, the dissolution questionnaire with tax document upload in Pennsylvania, the annual report and amendment changes in Texas, the compliance questionnaires in Arizona, the industry code (NAICS) for six products in West Virginia, the management pages for LLCs in Massachusetts, and the address and contact fields in the District of Columbia. In several of these tasks the requirements changed while the work was in progress. They also investigated production issues and support tickets.
Our Engineers Were Responsible for the Quality of Their Own Changes and Tested Them Before the Client's Acceptance
Our client's front end had to work flawlessly, because any mistake in a field would get the filing rejected by the state authorities or make the client's staff fix the order by hand, so our engineers were responsible for the quality of their own changes and thoroughly tested their code.
When our engineers added a new field or a new page, they had to write unit tests for it (for example, for the address checks in California and for the new dissolution pages in Pennsylvania) and update the client's automated tests.
They tested their changes in the client's QA environment before they were released to production. They created test orders to cover the edge cases, and for rules that apply to many states they verified the change for all of these states. Sometimes they found problems that were not in the ticket (for example, inconsistencies in the validation rules), and fixed them as well.
To prepare a change for the client's acceptance (UAT), our engineers had to test and check it, write clear instructions and attach screenshots, so the client's team knew what to check before release.