AI Hotel Call Auditing

Outcome

Reduced a documented 30–60 minute manual review workflow to 2–5 minutes.

Problem

Hotel-call review was slow and manual. Transcription alone was not enough; reviewers needed structured analysis they could use.

My role

I built the whole product alone: data storage, background processing, and the review interface.

Approach

I combined Django, Next.js, PostgreSQL, Celery, Redis, Whisper, and GPT-4. Background jobs handled transcription and model analysis so the web app stayed responsive. The app then presented the results in a form reviewers could use.

Lesson

Start with the slowest part of the work. A useful AI product also needs data storage, background jobs, a review flow, and clear failure handling.