Building a Callable Voice Agent | Dataford Interview Questions

Tell me about your approach to building a voice AI agent that users can actually call.

Implement the transcript-processing component for that agent. Given timestamped ASR segments, select the highest-confidence non-overlapping segments independently for each speaker, then merge adjacent turns from the same speaker when their silence is within max_silence.

Implement build_call_turns(segments, max_silence). Each segment is a dictionary with numeric start, end, and confidence, plus string speaker and text. Return chronologically ordered dictionaries containing only speaker, start, end, and text.

Constraints

0 <= len(segments) <= 500
0 <= segment.start < segment.end
0 <= segment.confidence <= 1
segment.speaker and segment.text are strings
max_silence >= 0
Segments with the same speaker may overlap; segments with different speakers may also overlap

Function Signature

def build_call_turns(segments, max_silence):

Your question is Building a Callable Voice Agent. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.

Problem

Tell me about your approach to building a voice AI agent that users can actually call.

Constraints

0 <= len(segments) <= 500
0 <= segment.start < segment.end
0 <= segment.confidence <= 1
segment.speaker and segment.text are strings
max_silence >= 0
Segments with the same speaker may overlap; segments with different speakers may also overlap

Function Signature

def build_call_turns(segments, max_silence):

Your question is Building a Callable Voice Agent. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.