Open to new projects Remote · GMT+5 · relocation

Launched a voice AI product line from zero · 80 paying B2B customers

Product Manager.

B2B SaaS · Voice AI agents · LLM

I launched Zvonobot AI, a voice AI agent line, from zero: 80 paying B2B customers and 500,000+ minutes of live conversations in production in under a year. I own pricing, unit economics and go-to-market, and I prototype with AI coding agents in days so hypotheses get tested, not debated.
Open to Product Manager roles in B2B SaaS, payments and AI products. Remote (GMT+5), open to relocation.

Message me on Telegram See the case

Currently

What I do now

Role: Product Manager, Zvonobot (Prof-IT Group), since Feb 2022. Launched the AI agent line in fall 2025 and own it since.
Team: 4 engineers; a sales team that grew from 4 to 15 to sell the line. Pricing and margins are mine.
Stack: LLM orchestration, ASR/TTS, SIP telephony, per-second billing, wallets, LLM call analytics

Projects

Selected work

Two tiers: the products I own at Prof-IT, and side projects taken to release in days.

01 · Day job

Products I own

Here I own product logic, metrics, and end-to-end delivery.

Flagship · 2026 Live in production

Zvonobot AI

Voice AI-agent platform for B2B (Prof-IT)

Product, pricing, go-to-market · launched from zero in fall 2025

Multi-tenant SaaS
Voice + LLM orchestration
B2B telephony

A Prof-IT product line I launched from zero in the fall of 2025 and own end to end: product, pricing, go-to-market. A multi-tenant SaaS console on top of telephony infrastructure: clients assemble and publish voice AI agents for outbound calls, from sales to reactivation and surveys.

What changed

80 paying B2B customers in under a year from launch.
The sales team grew from 4 to 15 to sell the line.
Average customer spend grew 30% after the AI plans launched.
30% of the line’s revenue comes from repeat payments.

Read the full case

cursivo

Live

A personal product I own from product logic to release. A CRM for car rental businesses covering the full cycle: fleet and utilization, bookings and contracts, customer base. AI handles the busywork: it reads and verifies client documents, scores deal risk, and suggests the next step to the manager. The goal: let a small rental operate like a chain, without growing headcount.

Type AI-powered car rental CRM

AI CRM
Car rental
In development

Read case Open cursivo.xyz

obrabot

Internal Prof-IT product

Internal Prof-IT product: a personal cabinet with analytics and dashboards for voice AI agents. AI segments calls and surfaces what needs human attention. I own the product.

Type Internal voice AI analytics dashboard at Prof-IT

LLM call analytics
Internal product
Prof-IT

Read case

02 · Built with AI coding agents

Side projects, shipped

Each one built personally with AI coding agents (Claude Code, Codex) in 1-7 days from idea to a working release. That is how I validate hypotheses and speak the same language as engineers. Each notes what it proves.

~2 days

bazabot

Telegram bot for cleaning client bases

Proves AI/ops in a real workflow: kills the manual base cleanup before call campaigns and speeds up data prep for the team.

Python · Telegram Bot API · CSV/XLSX · p1sms API

3 days

Zodial

Personalized PDF report platform

Proves A full product in 3 days: payments, delivery, and analytics end-to-end. Proof I can take a hypothesis to a working paid product.

Next.js · TypeScript · Prisma · SQLite · YooKassa · Resend

~7 days

Slopiq

iOS app for ski and snowboard sessions

Proves Product taste on a new platform: a native iOS app with an AI assistant, from zero to a working build in a week.

SwiftUI · iOS · Weather · AI Assistant

~2 days

portfo

This site

Proves This very site: the only public repo here, maintained agent-native through PRs. Proof of the approach in action.

Astro 5 · TypeScript · MDX · GitHub Actions

github →

How I test hypotheses fast · Claude Code & Codex

A hypothesis gets a release in days, not a sprint.

Prototypes and side projects I build myself with AI coding agents. A hypothesis gets a working release in days instead of a sprint, and I speak the same language as engineers. Product decisions, specs and review stay with me; this site runs the same way, every change an agent-written PR that I review.

Spec

I frame the problem, the constraints, and what “done” means.
Agents build

Claude Code and Codex write the code, run the checks, and open a pull request.
Review and ship

I read every diff myself and merge only what I can defend. A merged PR deploys to production.