01 / Conversational AI

AI Voice Interview Agent

A voice interview workflow that turns a conversation into a structured summary, evaluation flags, and a human review.

MY ROLE
Builder · ownership details pending
CONTEXT
Anonymized project · demo reconstruction
  • Conversational AI
  • LLM summaries
  • Human review
Interview / ReviewSample session 004
TRANSCRIPT EXCERPTInterviewer

Tell me about a difficult product decision.

Sample candidate

We narrowed the first release to one workflow.

REVIEW QUEUEEvidence before evaluation.

Decision described. Outcome needs a follow-up.

Human review required
Demo reconstruction · fictional transcript

01The problem

A conversation is difficult to review when the transcript, summary, and evaluation notes live apart. This project brings the interview workflow and human review into one place.

Evidence to addAdd the original user problem, who the users were, and one anonymized research note.

02What I owned

I built an AI voice interview project. The scope described here includes interview workflow design, LLM summaries, evaluation flags, and human review. My individual contribution within that scope still needs to be documented.

Evidence to addSpecify the flows, code, prompts, or integrations personally owned, and what teammates owned.

03Product decisions

Reconstruction choices, not a historical decision log:

  • Keep the transcript beside the review notes so a reviewer can inspect the evidence.
  • Flag missing context for a person to review instead of presenting an unsupported conclusion.
  • Make follow-up questions visible rather than hiding uncertainty inside a score.
Evidence to addReplace these reconstruction notes with one actual decision, the alternative considered, and why it was rejected.

04Technical approach

  • Conversational AI for the interview interaction.
  • LLM-generated summaries and evaluation flags.
  • A human review stage for interpreting the output.

Speech services, model provider, frontend, backend, storage, and deployment stack are not yet confirmed.

Evidence to addAdd the actual stack and a sanitized architecture diagram; include latency, consent, and retention decisions if applicable.

05What shipped

The described product scope is a voice interview workflow with summaries, flags, and review. Release status and the exact production feature set have not been supplied.

Evidence to addAdd a dated release, an approved walkthrough, or a sanitized screenshot. Distinguish prototype work from production work.

06Outcomes & lessons

No usage, evaluation quality, or time-saving claims are included without evidence.

Evidence to addAdd one observed user response or measured result, then one thing that changed after feedback.