HR reached out to me on LinkedIn, and after the initial screening I was given a take-home exercise. It involved using their platform to build a voice-to-voice agent. The platform itself was straightforward to use, though the documentation could be improved. The most time-consuming part was building a web server to wrap a legacy system, along with documenting the whole solution. My submission wasn't approved due to a lack of observability in the delivered agent — though the app did log information to the console and tool executions were logged in the web server, so I was a bit surprised by that feedback. It seems the reviewer evaluated mainly based on the recorded demo, without going through the code or documentation in detail. My takeaway for others: make sure your demo is polished and self-contained, since it may end up carrying most of the weight in the evaluation.