The visible feature was only the surface
Deshi Mula company pages surface community conversation, but job seekers need structured identity, salary, workplace, story, and cited-research context at the decision point.
The system also had to respect concrete constraints. Salary/workplace reports may be community-submitted and require independent verification. The extension depends on the hosted b4join API and cannot operate as a complete offline research system. Page matching and injection must remain narrow to avoid collecting unrelated browsing data.
My role covered product design and browser-extension engineering, with responsibility for browser integration, provider workflows, and privacy.
I turned the constraints into boundaries
Deshi Mula page detection/content script; injected research panel UI; typed thin client to hosted b4join endpoints; release-built Manifest V3 package plus public documentation site.
Keep the extension thin and exclude backend keys/research datasets, trading offline independence for a smaller distributable privacy surface.
Attach context only to recognized company pages instead of broad browsing surveillance.
Preserve source citations and caveats rather than presenting community data as authoritative.
The key decision: Keep the extension thin and exclude backend keys/research datasets.
The implementation had to prove the model
I delivered the work across the full path: Implemented company-page detection and panel injection. Built identity, workplace, salary, job, story, and cited-research sections. Integrated hosted b4join lookup endpoints and error/loading states. Produced release packaging, store artwork, privacy docs, ADRs, and checks.
The result is concrete: Users can open workplace context without leaving the company page. Inspect salary and narrative signals alongside source links. Install a packaged release. Understand that evidence is advisory and API-dependent.
The senior engineering lesson was that solving the visible workflow is only half the job. The architecture must also make constraints, failure modes, evidence, and ownership explicit enough for the next change to remain safe.