Petr Malyukov, CEO and Co-Founder of dTelecom – Interview Series

Petr Malyukov, CEO and Co-Founder of dTelecom leads the company’s mission to build the decentralized infrastructure layer for real-time communication and voice AI. Since co-founding dTelecom in 2022, he has driven the development of the Decentralized Real-Time Communication (DRTC) Network, a DePIN-powered stack built on Solana that delivers scalable voice, video, chat, and AI speech capabilities at a fraction of traditional centralized costs. His background in building AI-enabled communication platforms, including the Product Hunt #1 app YOUS with built-in AI translation, informs dTelecom’s focus on AI-native interaction, autonomous agents, and micropayment-based access to speech intelligence.

dTelecom is an open, decentralized real-time communication network designed to give developers and enterprises full ownership of their communication infrastructure. Built on Solana using a DePIN architecture, the platform provides a complete stack including RTC APIs for audio, video, and live streaming; a dual-engine AI Speech-to-Text API (Parakeet-TDT and Whisper) supporting 99+ languages with smart routing, VAD, noise reduction, and hallucination filtering; and ready-to-use applications like dMeet. By replacing centralized providers with a distributed node network, dTelecom enables up to 95% cost savings while maintaining performance, reliability, and security, positioning itself at the intersection of decentralized infrastructure, AI voice interfaces, and the emerging agentic economy.

You have spent more than 17 years building companies across telecommunications, Web3, and AI. How did that background ultimately lead you to found dTelecom, and what experiences convinced you that real-time communication needed a fundamentally different architecture?

Earlier in my career, I built several communication products with my current CTO, Vadim Filimonov. We’ve always split responsibilities the same way: I focus on product and strategy, while he goes deep on telecom engineering, shaped by his experience at companies like Motorola.

In 2019, we launched a communications app with a built-in AI translator, and that project made the infrastructure’s pain points very tangible. Latency, reliability, routing, and cost are what users feel immediately. In this environment, when you rely on someone else’s stack, you inherit their flaws, and that puts a ceiling on how far you can take the product.

At some point, we realized we were solving the same problems repeatedly, just at different layers. You can optimize UI, improve codecs, and fine-tune workflows, but if the underlying infrastructure is centralized and closed, you’re still boxed in by how that provider runs capacity, pricing, and integrations. That’s what pushed us to start dTelecom in early 2022.

We built an architecture where real-time communication runs on independent nodes that carry the media traffic, while blockchain stands for the coordination and accountability layer. We use it to register nodes, track performance, and enforce quality rules automatically.

The real-time communication (RTC) market is still dominated by centralized platforms. From your perspective, what structural limitations in today’s RTC systems made decentralization inevitable?

I wouldn’t call decentralization inevitable. Centralized RTC works, and it’s been good enough to build Zoom-, Twilio-, Agora-style products for years. The limitation appears when real-time communication becomes a core business layer, because control and accountability start to matter as much as audio quality.

Opacity is the key issue. In a centralized stack, users rarely know where traffic is routed, who touches the data plane, or how policies are enforced across regions. That creates a trust gap for privacy-sensitive use cases and concentrates risk, since a small number of infrastructure owners become aggregation points for metadata and operational logs. 

Besides, it caps flexibility, as jurisdiction-aware routing, auditability, and deeper customization remain tied to one provider’s roadmap.

Decentralization becomes compelling when it returns control to the edge. A node-based model lets developers choose where communication runs and makes performance auditable, so nodes register, uptime is measured, reputation is earned, and failures are recorded with clear accountability. Cost is part of it too, since you’re no longer paying a centralized provider’s margin and overhead.

dTelecom is built on Solana and designed around decentralized physical infrastructure networks (DePIN). Why was this architecture essential for supporting high-performance voice and video communication at scale?

We built on Solana and designed dTelecom as a DePIN network because high-performance voice and video depend on fast coordination. In RTC, latency is everything, so you need a chain that can keep up when you’re coordinating a global fleet of edge nodes and settling micro-payments in real time.

Solana has become the main “home” for DePIN due to its large active node base (230K+) and a mature ecosystem alongside throughput (around 4,000–6,000 TPS) and fast finality. All of this makes it possible to coordinate node activity quickly enough for real-time communication, without adding the kind of friction that would slow the network down.

Additionally, communication stops being a black box, expensive utility, and becomes something transparent and measurable, where algorithmic accounting can support SLA enforcement at a Tier-1 level. Over time, that creates a flywheel: better tooling, liquidity, and AI agents concentrate in the same place, which makes scale easier to achieve than in traditional Web2 models.

Artificial intelligence plays a central role in your network. How is AI currently used within dTelecom to manage routing, performance optimization, automation, or security?

At our company, AI is part of how the network is managed day to day. We use predictive routing, where algorithms continuously monitor the condition of hundreds of decentralized nodes and can anticipate connection degradation before the user notices. That lets the system switch streams to the most stable nodes in real time, aiming for 99.9% uptime in a distributed environment where there’s no single point of failure.

In parallel, we’re building toward “AI at the edge.” Instead of pushing raw, heavy media data around the network, AI runs on the node level to optimize media streams directly — everything from adaptive compression to neural noise suppression on the fly. In effect, the network transports an optimized “intelligence layer,” which reduces load on bandwidth and makes communication faster and cheaper.

More broadly, this is how we see quality RTC scaling without costs exploding. The goal is to turn infrastructure from a passive pipe for traffic into an active, intelligent system that helps maintain quality and protect data privacy by design. We’ve also integrated the x402 protocol, so voice-native agents can tap into our STT on demand, pay $0.005 per minute, and convert voice to text to complete tasks. In practice, this creates efficient AI rails for agent-to-human communication at scale.

Cost reduction is a major theme in your messaging. What specific design decisions allow dTelecom to operate more efficiently than traditional centralized communication providers?

I’d say the most significant design decision is the infrastructure model itself. We built dTelecom as a shared-cost network, where RTC workloads are distributed across community-operated nodes, rather than carried by one company’s global server farms the way traditional providers do.

Specifically, this changes the cost structure in three ways. First, we avoid the fixed overhead that comes with owning and operating a centralized fleet, from server maintenance to bandwidth provisioning. Second, scaling costs drop because capacity can expand through the node network during peak demand. Ultimately, reliance on third-party cloud providers is lower, which reduces layered markups and makes pricing more predictable as usage grows.

You can see this cost advantage in our x402-powered speech-to-text. At $0.005 per minute, dTelecom STT undercuts OpenAI Whisper ($0.006), Deepgram Nova-3 ($0.0077), Google Cloud STT ($0.016), Azure Real-time ($0.0167), and AWS Transcribe ($0.024) — while agents can pay programmatically per minute and run tasks without accounts or API keys.

That way, we ensure lower entry barriers and more stable economics for startups, SMEs, and emerging projects, without trading off quality or performance.

Privacy and data control are growing concerns for enterprises relying on live voice and video platforms. How does a decentralized approach change the way communication data is handled and protected?

A decentralized model changes the default data posture. In centralized RTC, a lot of communication data ends up flowing through a small number of infrastructure owners and centralized storage points, which concentrates risk. A peer-to-peer, node-based approach, helps to remove those single aggregation points, which can significantly reduce the blast radius of a large-scale breach.

In our case, communication data is handled locally and securely, with end-to-end encryption. The architecture also supports jurisdiction-aware routing and optional on-prem deployment patterns, so enterprises have more control over where data goes and how it’s processed.

And from a compliance perspective, the infrastructure can help enterprises align with frameworks like GDPR and CCPA, even though the final compliance outcome still depends on how each company integrates the technology and what their specific legal obligations are.

dTelecom has received grants and recognition from organizations such as Google, the Solana Foundation, and ElevenLabs. How did that early validation influence your execution priorities as a founder?

Support from Google Cloud, Solana, and ElevenLabs was the moment that took us out of “let’s prove the idea” mode and into “we’re being judged like infrastructure” mode. Once you get that kind of validation, you have to justify it in action by delivering reliability at a Tier-1 level.

Primarily, this endorsement pushed us beyond being just a “pipe” for traffic and toward an intelligent platform, where voice AI is integrated directly into the communication layer.

Our main priority has become execution speed and developer experience, as we’re putting a strong focus on developing a production-ready SDK that allows teams to integrate our solution in minutes. That way, we make decentralization “invisible” and convenient, converting trust from giants into real traction and turning the AI and RTC crossover into our key competitive edge.

As decentralized communication networks scale, what technical challenges do you believe will matter most, particularly around latency, coordination, and reliability?

The way I see it, scaling real-time DePIN networks is basically a fight for standardization in a controlled kind of chaos. The key challenge is guaranteeing quality (SLA) in a trustless environment, where you don’t own the physical infrastructure. In dTelecom, we address that through algorithmic reputation, since for us, reliability is a matter of math — we’re building a system where honest node performance is economically more beneficial.

Technically, the critical point becomes synchronizing the state of thousands of nodes around the world. To “conduct” them without delays, we use Solana as a high-speed transaction layer, though most of the heavy lifting happens at the level of optimized P2P protocols that let nodes exchange data almost at the speed of light.

And then there’s the challenge of heterogeneous hardware: every node has different bandwidth and capacity. In response, we deploy an adaptive abstraction layer. After all, the future belongs to protocols that adjust on the fly to a node’s capabilities, smoothing out heterogeneity and turning a mixed network into a stable infrastructure layer.

You emphasize infrastructure ownership as a core advantage of decentralized communication. How does this shift reshape the business models available to developers and platforms building on top of dTelecom?

In Web2, developers pay a lifetime “cloud tax” that scales with their business. In our co-ownership model, infrastructure shifts from a passive expense (OPEX) into a strategic asset, so platforms can run their own nodes, recoup part of the cost, and reinvest into their network.

Removing intermediaries like Twilio or Agora — who can take up to 80% of the margin — changes unit economics dramatically. It lets developers make viable products that otherwise would be hard to justify: offering free, enterprise-grade basics (like dMeet, our open-source video conferencing app), or giving creators an unusually high revenue share (like FROGY, our Telegram-based live-streaming mini-app) while still being profitable.

It also opens the door to “programmable communication.” When you have infrastructure ownership, you get deeper control and access to “raw” data that cloud services typically keep closed. So developers can plug their own AI models directly into communication nodes and build a new class of applications — from ultra-private messengers to global streaming networks with entirely different economics.

Looking ahead, how do you expect decentralized, AI-native communication infrastructure to change how people and businesses interact online over the next decade?

Over the next decade, I expect AI-native communication to become an inherent layer in every digital product. Voice, video, and real-time interaction will be everywhere — in support, education, commerce, creator platforms — and the expectation will be that it just works, instantly, at global scale.

With dTelecom, I strive to make this foundation faster, more reliable, and genuinely user-owned, while still being scalable and enterprise-ready. So if we get it right, real-time communication becomes cheaper to build, easier to trust, and more useful in everyday life — without forcing everyone to rent the same centralized stack forever.

Thank you for the great interview, readers who wish to learn more should visit dTelecom.