About
Experienced Chief Executive Officer with a demonstrated history of working in the…
Articles by Alexey
Activity
6K followers
-
Alexey Aylarov reposted thisAlexey Aylarov reposted thisBuilding voice agents on JEV by TypeSafe AI and Voximplant is becoming the new normal Georgee Lomidze published a practical breakdown of how his team at PocOS made an AI phone receptionist fast and predictable Voximplant runs the call and the real-time layer, Gemini Live handles speech, and JEV judges the consequential moments (did the caller confirm the booking, did they say goodbye) while plain code makes the final decision Worth the 14 minutes if youre taking voice AI to production: https://lnkd.in/e-G5-n8z #VoiceAI #Voximplant #AIAgents #jev #typesafeaiJEV in Production: How We Turned Our AI Phone Assistant into a Fast, Predictable, JEV-Gated SystemJEV in Production: How We Turned Our AI Phone Assistant into a Fast, Predictable, JEV-Gated System
-
Alexey Aylarov shared thisTrue speech-to-speech models and APIs are here to stay. Use Voximplant's GPT Live API Integration to launch your Voice AI agents in full-duplex mode and carry less of the burden when it comes to things like turn detection and response latency.Alexey Aylarov shared thisGPT-Live-1 by OpenAI now runs on Voximplant Platform We got used to voice agents freezing on hard questions. Ask for a lookup or anything that needs real reasoning, and most go quiet mid-call, then dump the whole answer back in one block once they're done thinking. But with the new OpenAI model, that doesn't happen. GPT-Live-1 keeps talking to you while working on the question in the background, then brings the answer back into the conversation when it's ready. This makes the whole experience a lot smoother It's also noticeably better at catching order numbers and confirmation codes correctly the first time. On longer calls, it holds onto what was said earlier instead of making the caller constantly repeat themselves A smart model is only half the job though, it still needs to run on a real phone conversation. On Voximplant, GPT-Live-1 is already wired into phone, SIP, WebRTC and WhatsApp, so you can go test this out right now https://lnkd.in/e7wexrr5
-
Alexey Aylarov shared thisGoogle has released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, and both are now available to Voximplant developers. Simply set the Gemini Live API integration to gemini-3.8-live or gemini-3.8-live-extended-thinking, and you’re ready to go.Alexey Aylarov shared thisGemini 3.8 Live and Gemini 3.8 Live Extended Thinking launched today, and both are already running on Voximplant Live is built for tight conversational latency, so the model responds inside a live call without the lag that breaks a conversation, the kind of gap that makes a caller think the line dropped Extended Thinking trades some of that speed for background reasoning: it keeps working and calling tools between turns instead of going quiet, which matters for anyone building agents that check a database or hit an API mid-call If you're already building on our platform, it's now available to you. And if you're not with us yet, this is a pretty good excuse to give Voximplant a try Huge congrats to the Gemini team at Google on the launch
-
Alexey Aylarov posted thisVoximplant : 390M+ calls connected successfully in August alone. Scaling our Voice AI backbone to power conversations worldwide.🚀
-
Alexey Aylarov reposted thisAlexey Aylarov reposted thisVoice AI pilots stall for one reason more than any other: the client needs compliance answers covering every provider in the stack before they can proceed Every component in a voice AI pipeline is a potential data transfer. Each one needs an answer, and if you can't give it for even one provider, the conversation stops there Voximplant gives you the tools and infrastructure to answer those questions across the stack. The telephony layer supports EU data residency and comes with compliance documentation. For LLM layer we added baseUrl support to give you control over where the requests go Here’s a full guide to running a sovereign voice AI stack on Voximplant: https://lnkd.in/d-Z5Xgfs Not sure if Voximplant fits your compliance requirements? Reach out, we are happy to help
-
Alexey Aylarov posted thisVoximplant now handles more than 300 million successful calls / month. If you need Voice AI at scale - our serverless platform is here to help, give it a try today! 🚀
-
Alexey Aylarov shared thisThe fastest and easiest way to connect your calls to Inworld AI Realtime API - Voximplant's new integration with it! Whether your users are calling in via a standard phone line, SIP, or WhatsApp, Voximplant now has you covered with a seamless integration that gets you live in minutes. See it in action https://lnkd.in/e35JSqqPConnect Inworld Realtime AI Agents to Phone Calls & SIP (Voximplant)Connect Inworld Realtime AI Agents to Phone Calls & SIP (Voximplant)
-
Alexey Aylarov reposted thisAlexey Aylarov reposted thisWe’ve published Voximplant AI Agent Skills Now agents like Cursor, Claude Code, or Codex can write VoxEngine scenarios, automate platform setup, deploy and test call flows, and debug real calls from logs This is part of a broader direction for us: making Voximplant agents-friendly. We believe that as AI coding tools become a standard part of how developers work, the platforms they build on should work just as naturally with those tools All you need is a Voximplant account and a service account for access. Your AI agent will walk you through the setup Try it out and share your feedback: https://lnkd.in/dFJQnNiD
-
Alexey Aylarov shared thisIntegrating your Voice AI with other tools and services? The MCP client is ready to help. You may not always need it, but when you do, Voximplant has your back.Alexey Aylarov shared thisVoice AI infrastructure should give developers control, not take it away Today we’re shipping a native MCP Client in VoxEngine. Connect to any MCP server directly from your scenario code. No gateway, no middleware, no extra service to maintain. The same tool layer works across providers: switch models without rebuilding your integrations You can now have full visibility over every tool call, the ability to add telephony context, adapt parameters, handle fallbacks — all in the same place you manage the rest of the call Demo: watch an agent create a live Zapier request before the caller hangs up → https://lnkd.in/dJ3KvDQf Docs: check the setup guide, architecture, and full scenario code → https://lnkd.in/d5SwK4SB Blog post: read more → https://lnkd.in/d82ZwidC
Alexey Aylarov replied to a comment
2d
Maksim Barkov they have howto for that, basically you create pre-playback buffer and make your checks before playing back media to the user. It adds some complexity, but if it's important it's worth it.
-
Alexey Aylarov liked thisAlexey Aylarov liked thisWe’re excited to announce that Ultravox is now part of Inworld AI. We’ve worked closely with Inworld for over a year through our text-to-speech partnership. We share a focus on real-time AI at scale, and that collaboration showed us how much more we could build together. At Ultravox, we’ve focused on helping developers build voice agents that understand people, reason through requests, and get things done: with the turn-taking and interruption handling that live conversations demand. Inworld brings speech models, model serving, and real-time inference infrastructure. Together, we can improve the full conversation: understanding what someone says, helping them accomplish a task, and responding in a voice that fits the moment. Key members of our team have joined Inworld to keep developing Ultravox. For the developers and customers building with us, your agents keep running, and the first improvement is already live. Every built-in Inworld voice on Ultravox is now upgraded to Realtime TTS-2. Your existing voice IDs stay the same, with no code changes required and no additional cost. Thank you to the developers, customers, and community who have helped shape Ultravox. We’re excited to keep building with you, with more to come in the next few weeks. Read more about why we came together, what you get today, and what’s next in the comments.
-
Alexey Aylarov liked thisAlexey Aylarov liked thisWe’re excited to announce that Ultravox.ai is now part of Inworld. Ultravox is the platform developers use to build real-time voice agents. We've worked with the team for a while through our TTS partnership, and today members of the team that built it are joining Inworld to keep developing it. Ultravox brings speech understanding, reasoning and task completion, along with turn-taking and interruption handling in live conversations. Inworld brings the speech models, model serving and real-time inference underneath. Together, we can improve the full conversation: understanding what a user says, helping them get something done, and responding in a voice that fits the moment. The first improvement is live today. Every built-in Inworld voice on Ultravox now runs on Realtime TTS-2, with existing voice IDs unchanged, no code changes required and no additional cost. This is a big step in our work on speech-to-speech experiences that bring speech understanding, reasoning, and expressive voice generation closer together, and we're only getting started. Full post on why we came together, what you get today, and what's next: https://lnkd.in/g9qZJ4Uz
-
Alexey Aylarov liked thisIt’s my birthday today. Coincidence? 😀 Welcome to the team, Ultravox.ai! Go Inworld AI! 🚀Alexey Aylarov liked thisWe’re excited to announce that Ultravox.ai is now part of Inworld. Ultravox is the platform developers use to build real-time voice agents. We've worked with the team for a while through our TTS partnership, and today members of the team that built it are joining Inworld to keep developing it. Ultravox brings speech understanding, reasoning and task completion, along with turn-taking and interruption handling in live conversations. Inworld brings the speech models, model serving and real-time inference underneath. Together, we can improve the full conversation: understanding what a user says, helping them get something done, and responding in a voice that fits the moment. The first improvement is live today. Every built-in Inworld voice on Ultravox now runs on Realtime TTS-2, with existing voice IDs unchanged, no code changes required and no additional cost. This is a big step in our work on speech-to-speech experiences that bring speech understanding, reasoning, and expressive voice generation closer together, and we're only getting started. Full post on why we came together, what you get today, and what's next: https://lnkd.in/g9qZJ4Uz
-
Alexey Aylarov liked thisI almost didn't join Inworld AI. When the Ultravox.ai acquisition started coming together, my first reaction was that Inworld was a TTS company. I respected the work, but I was seriously weighing something different and didn't see Inworld as a potential part of my next chapter. Then I sat down with Kylan Gibbs and the leadership team. What I heard wasn't a TTS roadmap. It was a plan to build the entire value chain for realtime AI: high scale, high quality, low latency, and low cost. In this industry you usually get to pick two, maybe three. Inworld is building for all four, because that's what it takes for consumer apps to bring the benefits of AI to everyone, not just the people who can pay for the premium tier. That's the problem I want to work on: bringing model research and infrastructure together so developers can put realtime AI into any app or experience. So today, as Inworld acquires Ultravox, I'm joining as Head of Developer Platform. That covers Ultravox alongside Inworld's TTS, STT, and LLM inference. Ultravox's strength in live conversation (understanding speech, turn-taking, handling interruptions) is going to accelerate Inworld's full speech-to-speech capabilities. I joined Fixie in early 2023 as founding head of developer relations. The job grew well past the title into pricing, partnerships, operations, and support, and I loved every stretch of it. One memory I'll always keep: testing our first realtime voice platform. We built a fictional drive-thru called Dr. Donut, and the team kept trying to trip it up. My favorite: "I'm in a hurry, can you dunk the donuts in coffee for me?" Listening to people try to break it, and the model mostly holding its own, never got old. Then there was the culture. We had an internal currency, and the sushi burrito was an official unit for measuring cost. As the person who also ran pricing, I stand by it as a benchmark :) To the Ultravox team: thank you for building something serious without ever taking yourselves too seriously. To everyone heading to Shopify, they're getting great people. To everyone coming with me to Inworld, let's keep going. Thank you to Zach Koch for his tireless commitment to our customers and our team, and to every developer and customer who bet on us early and grew with us. To Ultravox customers: the platform isn't going anywhere, and the first upgrade is already live. Every built-in Inworld voice now runs on Realtime TTS-2, with no code changes and no added cost. More in the coming weeks. And yes, Dr. Donut is still open.Alexey Aylarov liked thisInworld has acquired Ultravox.ai to accelerate our speech-to-speech efforts. We've worked closely with Ultravox for over a year through our text-to-speech partnership. We're both focused on realtime AI at scale, and we both care about improving everyday life with AI. Over time we saw how complementary our work is. Ultravox brings expertise in live conversation, like turn-taking and interruption handling. Together we can improve the whole speech-to-speech experience, from understanding what someone says to answering in a voice that fits the moment. The care that went into Ultravox shows the first time you use it. The team members who have joined Inworld will keep developing the platform. For Ultravox customers, the first improvement is live today. Every built-in Inworld voice is upgraded to Realtime TTS-2, with no code changes and no added cost. More to come in the next few weeks. Welcome to the Inworld family. Excited to build this together.
-
Alexey Aylarov liked thisI think this is big. 🌠 FastAPI's objective is to provide what you always need for real prod apps. And not provide what you don't. You always want metrics, there's just not enough time to always set it all up. Now it's just there, by default.Alexey Aylarov liked thisFastAPI now has built-in OpenTelemetry support, natively 📈 No extra code needed. See how your app behaves in production. Go and upgrade to "fastapi[standard]" version 0.142.2 ⬆️ Read the docs: https://lnkd.in/dPTg7VA7
-
Alexey Aylarov liked thisAlexey Aylarov liked thisBuilding voice agents on JEV by TypeSafe AI and Voximplant is becoming the new normal Georgee Lomidze published a practical breakdown of how his team at PocOS made an AI phone receptionist fast and predictable Voximplant runs the call and the real-time layer, Gemini Live handles speech, and JEV judges the consequential moments (did the caller confirm the booking, did they say goodbye) while plain code makes the final decision Worth the 14 minutes if youre taking voice AI to production: https://lnkd.in/e-G5-n8z #VoiceAI #Voximplant #AIAgents #jev #typesafeaiJEV in Production: How We Turned Our AI Phone Assistant into a Fast, Predictable, JEV-Gated SystemJEV in Production: How We Turned Our AI Phone Assistant into a Fast, Predictable, JEV-Gated System
-
Alexey Aylarov liked thisAlexey Aylarov liked thisGPT-Live-1 by OpenAI now runs on Voximplant Platform We got used to voice agents freezing on hard questions. Ask for a lookup or anything that needs real reasoning, and most go quiet mid-call, then dump the whole answer back in one block once they're done thinking. But with the new OpenAI model, that doesn't happen. GPT-Live-1 keeps talking to you while working on the question in the background, then brings the answer back into the conversation when it's ready. This makes the whole experience a lot smoother It's also noticeably better at catching order numbers and confirmation codes correctly the first time. On longer calls, it holds onto what was said earlier instead of making the caller constantly repeat themselves A smart model is only half the job though, it still needs to run on a real phone conversation. On Voximplant, GPT-Live-1 is already wired into phone, SIP, WebRTC and WhatsApp, so you can go test this out right now https://lnkd.in/e7wexrr5
Experience & Education
Languages
-
English
-
-
Russian
-
Other similar profiles
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content