मुख्य सामग्री पर जाएं

skills · Voice AI agents

Voice Agents & Conversational AI

Build a voice AI agent, and measure it from the first call.

कोर्स के अंत तक, आप ये कर पाएँगे:

  • A call spec with a wait target
  • A live agent with its timings logged
  • Recognition scored on key terms
  • An end-of-turn setting chosen from a test
  • Dates and ids said in spoken form
  • A tool that books only after a read-back
  • A handoff guard and test calls
  • A phone number, an honest opening and a cost per call

$9/माह से8 घंटे5 अध्याय20 पाठ

कोर्स अभी अंग्रेज़ी में हैं।

“Where did the caller's second and a half go?”

इस कोर्स के बारे में

यह किसके लिए है

  • You write Python and want to build a voice agent that answers real calls, not a demo
  • You have tried a voice demo and want to know why callers talk over it or hang up
  • You need an agent you can hand to a team: measured, tested, and honest with callers

आपको क्या चाहिए

  • You write Python and use git
  • A laptop with a microphone (a free LiveKit Cloud account for the live lessons)

आप कौन से कौशल अभ्यास करेंगे

  • Voice agents
  • LiveKit Agents
  • Speech-to-text
  • Text-to-speech
  • Turn detection
  • Function calling
  • Telephony (SIP)
  • Conversation testing

कितना समय लगता है

कुल मिलाकर लगभग 8 घंटे

  1. अध्याय 1 · A call is a race against silence2 घंटे
  2. अध्याय 2 · Hearing the words that matter1.5 घंटे
  3. अध्याय 3 · Speaking so people can follow1.5 घंटे
  4. अध्याय 4 · The conversation engine1.5 घंटे
  5. अध्याय 5 · Answering real calls1.5 घंटे

इसमें क्या शामिल है

  • 20 पाठ
  • 78 पठन
  • 20 हल किए गए उदाहरण
  • 40 अभ्यास
  • 40 त्वरित जाँचें
  • 20 आम भ्रम
  • 20 क्विज़
  • अध्याय के अंत की 5 जाँचें

पाठ्यक्रम

5 अध्याय, 20 पाठ

पूरा कोर्स: हर अध्याय और पाठ। विस्तार के लिए अध्याय पर टैप करें। शुरू करने पर पाठ खुलेंगे।

    • Measure What It Heard
    • When the Transcript Invents Words
    • Teach It Your Words
    • When Has the Caller Finished?
    • Write for the Ear
    • Choosing a Voice
    • When the Caller Talks Over It
    • Silence, Waiting and Goodbye
    • Tools That Act, and the Read-Back
    • The Steps It May Not Skip
    • Hand Off to a Person
    • Test Calls Before Real Callers
    • Put It on a Phone Number
    • Say It Is an AI, and Ask Before Recording
    • What a Call Costs
    • After Launch: Read the Calls

करियर

यह कोर्स आपको कहाँ तक ले जा सकता है

वे भूमिकाएँ जिनसे यह कोर्स जुड़ता है, और वे कौशल जिन्हें कोर्स पूरा करने के बाद आप अपने रिज़्यूमे और LinkedIn पर लिख सकते हैं।

वे कौशल जो आप बनाएँगे

कोर्स पूरा करने के बाद इन्हें अपने रिज़्यूमे या LinkedIn प्रोफ़ाइल में चिपकाएँ।

  • Reading voice agent latency by stageAnalyze end-to-end latency of a voice AI pipeline
  • Running a voice agent locally with LiveKit AgentsBuild voice AI agents in Python
  • Choosing between a cascaded pipeline and a speech-to-speech modelEvaluate STT-LLM-TTS pipelines against realtime speech models
  • Measuring time to first audio against a targetMeasure and report time to first token and time to first audio
  • Measuring speech recognition accuracy for a voice agentEvaluate speech-to-text accuracy with word error rate
  • Designing confirmation steps and speaker tests for voice agentsAdd confirmation (read-back) before a voice agent acts on recognized speech
  • Tuning speech recognition with key terms and controlled comparisonsConfigure keyterm prompting or custom vocabulary for speech-to-text
  • Configuring end-of-turn detection for a voice agentTune endpointing and turn detection for conversational AI
  • Normalizing text for speech in a voice agentConvert dates, numbers and identifiers into spoken form for text-to-speech
  • Evaluating and choosing a text-to-speech voiceEvaluate text-to-speech voices on latency, pronunciation and cost
  • Handling interruptions in a voice agentConfigure barge-in and interruption handling for a conversational agent
  • Designing waits, silence handling and call endingsDesign filler speech and idle handling for long-running tool calls
  • Designing voice agent tools with caller confirmationImplement function calling for a voice agent with confirmation before state-changing actions
  • Enforcing conversation flow outside the modelBuild deterministic guardrails and state machines around LLM tool calls
  • Designing human handoff for a voice agentDesign escalation and human handoff rules for a voice AI agent
  • Testing a voice agent with scripted callsWrite regression tests and text-mode evaluations for a conversational AI agent
  • Connecting a voice agent to a phone numberConnect a voice agent to telephony with SIP dispatch
  • AI disclosure and recording notices for voice agentsImplement AI disclosure and call recording notices in a voice agent
  • Costing a voice agent callModel the per-call cost of a voice agent across speech, model and telephony
  • Reviewing live calls and writing a launch noteMonitor tail latency and call outcomes for a voice agent

प्रश्न

प्रतिबद्ध होने से पहले

You write Python and have called an LLM API, and you want an agent that answers a call or a browser mic: hears the caller, answers fast enough, books things correctly and knows when to pass the call to a person.

Not to start. As of November 2026, LiveKit Cloud's free Build plan includes LiveKit Inference credit for the speech and model calls in the live lessons, and one free US phone number. Every lesson also runs on the course's invented sample with no account.

The course does not aim for that. Lesson 5.2 has your agent say it is an AI at the start of the call, as laws such as the EU AI Act require in many cases, and the agent hands calls to a person when it should.

No. You need Python. The course explains speech recognition, turn detection and speech synthesis as far as you need to measure and tune them.

Your own call makes the course worth more, and each lesson has one required step on it. If you have none, take the invented health-centre booking line, which has a path through every lesson.

आज ही शुरू करें