VerbalVictory

The brief

Reps get almost no structured feedback on their calls, and the recordings that could teach them sit unwatched. The product needed to turn a raw call into a fast, specific score plus coaching a rep would actually act on — without making them wait.

What was built

• A multi-pass scoring engine that returns a headline score in seconds, then layers in deeper per-dimension coaching and a narrative breakdown.
• An audio path that transcribes the call and feeds the scoring pipeline.
• A progressive UI that shows the fast score first and streams in the richer analysis so the user is never staring at a spinner.

What it demonstrates

Real-time LLM scoring with a staged-response architecture — the kind of latency-aware AI product work that separates a usable tool from a slow demo.