Replacing a broken data service with a tool that kept working

An HR analytics company relied on an outside service for LinkedIn profile information. When that service stopped working, an important part of the company’s process stopped with it. One or two development teams had already tried to build a replacement but had not delivered a working tool.

LinkedIn regularly changes how its pages are arranged. A tool that expected every name, job, and school to appear in one exact place might work today and fail tomorrow. The replacement had to understand the page more generally.

I built a Chrome extension that adds a collection button directly to LinkedIn. It visits the main profile and its detailed sections, gathers the useful text, links, and images, removes repetition, and reports whether the collection succeeded.

Instead of relying on the exact position of every field, the tool first captures the meaningful content and its relationships. A separate service cleans that material and uses AI to organize it into consistent fields such as work history, education, awards, languages, and publications. The final record is saved in cloud storage.

The collector does not need to know the company’s final data format, and the organizer does not need LinkedIn to keep the same design forever. That separation is what makes the tool more durable.

The company no longer had to wait for an outside vendor to restore the service. Its staff could run the tool directly, see whether it worked, and trace each saved record back to its source and collection time. I continued improving its coverage, speed, error handling, and user feedback after the first delivery.

LinkedIn has changed its page templates since the tool was delivered, but the system has continued working. The company recovered the capability it had lost without returning to the failed vendor.

The project began with a broken outside service and two unsuccessful replacement attempts. It ended with a working internal tool that has continued doing its job.