Apr 20, 2026
Inheriting a production app
When I joined Teams Squared, I inherited an internal security reporting tool and an unreleased client dashboard. We set aside the other prototypes to focus on these two applications.
The dashboard had no active users yet. I needed to understand the existing code, work out what the company needed from it, and prepare it for release as the sole developer.
Expanding the client portal
The original app let clients view team members, invoices, and notifications. It pulled timesheets and leave requests from Zelt and Clockify through APIs and scraping.
We decided to bring those workflows into the app instead of continuing to depend on the external tools. Clients could manage their account, billing, and team roster, while contractors could log hours, manage timesheets, and request leave. We added Resend for transactional emails and configured Microsoft Entra ID single sign-on, which was staged for production release at the time of writing.
What I missed in the first version
At first, I focused on getting features working. I added timesheet approvals and leave tracking, but spent too little time checking how users would move through the app.
Some features worked as specified but were confusing to use. If I were starting again, I would map out the daily workflows with clients and contractors before building the screens. That would help me catch navigation and usability problems earlier.
Supporting the rollout
We did not have a dedicated QA team. When clients and contractors started using the app, I was responsible for investigating bugs, fixing them, and deploying updates.
I used Claude Code and Google Antigravity to help trace issues across the React front end and FastAPI back end and draft fixes. I also separated staging from production, set up automated deployment pipelines, and used AI tools to help write regression tests and check builds before release.
What I would carry into the next project
I would spend more time understanding existing code and watching how people use the application before expanding its feature set. The rollout showed me where my assumptions about users were wrong, and staging and regression checks helped me make corrections more safely.