Description: We're looking for an experienced software developer to build a custom AI voice agent that can conduct primary research interviews with experts sourced through networks like GLG, Third Bridge, AlphaSights, and similar platforms. What the tool needs to do: The agent should accept an uploaded interview guide (structured discussion outline with key questions and topics), then autonomously conduct a live phone/video call with a human expert — working through the guide while adapting in real time. This means it can't just read questions off a script; it needs to listen to responses, ask intelligent follow-up and probing questions, manage conversational flow naturally, and ensure all key topics from the guide are covered before wrapping up. Core requirements: Upload and parse interview guides (PDF, Word, or similar formats) to extract questions, topics, and probing areas Real-time voice conversation with a live human participant — natural-sounding speech, low latency, and good turn-taking Dynamic follow-up questioning — the agent should probe deeper on interesting or vague answers rather than rigidly moving to the next question Topic tracking — awareness of which guide topics have been covered and which remain, with graceful transitions Call recording and transcription with structured output (transcript, summary, key findings mapped back to the guide) Integration with standard calling infrastructure (VoIP/telephony so it can dial into or receive calls) Nice to have: Configurable interview "persona" or style (e.g., tone, level of formality, domain context) Ability to flag contradictions or noteworthy quotes in real time Dashboard or simple UI for uploading guides, launching calls, and reviewing results Support for multiple concurrent interviews About you: Strong experience building voice AI applications (using tools like OpenAI Realtime API, Vapi, Bland.ai, LiveKit, Deepgram, ElevenLabs, or similar) Familiarity with LLM orchestration for multi-turn, goal-directed conversations Experience with telephony integration (Twilio, SIP, etc.) Bonus: understanding of management consulting, private equity, or investment research workflows where expert calls are common Please share relevant past projects or demos in your proposal. We're open to discussing tech stack choices — what matters most is a reliable, natural-sounding agent that can handle a 30–60 minute semi-structured interview without falling apart.