61M+ installs, company-reported
Nicegram GPL-2.0
Privacy-first Telegram client, open source on iOS, Android and desktop. 61M+ installs, as reported by Nicegram on 24 June 2026.
siarhei@svlab ~ % whoami
Siarhei Sheleh
Technical entrepreneur · products used by millions
Technical entrepreneur in Warsaw, Poland. 13 years building and launching products used by millions of people — mobile, desktop, web. I stay in the technical work: AI integrations and agent systems, marketing and research tooling, growth and retention mechanics, and the go-to-market that puts it all in front of users. SV Lab is where I ship it.
- Product engineering
- AI integrations
- Agent systems
- Marketing & research tools
- Growth & retention
- Go-to-market
// NOW: making agents build products you can ship, scale, and maintain — not neuroslop.
// YEARS 13+ shipping
// SHIPPED 10 products
// LIVE 6 running
// BASE Warsaw, PL
01 About me
One person accountable, all the way down.
I'm Siarhei Sheleh. I build, market, and operate products from Warsaw, and I'm the one in the repo: no account manager between you and the work, no handoff to a bench you never met.
What I own end to end is the whole surface a product needs to exist: research, engineering, go-to-market, marketing, and operations. Most orgs split these across five roles and pay for the handoffs. I run them as one loop — and ship faster because of it. That loop is what 10 shipped products and every consulting engagement run on.
02 Open source
Shipped in the open,
used by millions.
Two lines of public work. The Nicegram platform — an open-source client tens of millions of people run every day, plus the business layer now entering early access. And sshlg-skills, the skill family that teaches coding agents how I take an idea to production.
Nicegram Business EARLY ACCESS
Nicegram OS — the AI business layer for Telegram-native teams. Company-owned accounts, delegated access, retained chat history, approved AI agents with a controlled context layer, and CRM sync. Revoke a seat and the deals stay with the company.
sshlg-skills
MIT 6 skills · 13 years of product judgementThe way I take an idea to production — discovery, gated delivery, product marketing, onboarding, retention, search — written down as skills a coding agent follows.
-
/01 task-pipeline
Intake grill that refuses vague work, then nine gated stages from docs and spec to deploy and post-deploy checks.
-
/02 super-ux
Scenario-driven UX: personas and jobs, flows, screens, scenarios, audits — plus a linter that catches docs drifting from the code.
-
/03 seo-aeo-audit
Evidence-first SEO and AEO/GEO audit: ten tracks, evidence tiers, a dated Google update timeline, and 59 prioritized growth plays.
-
/04 NEW agent-sync
Several agents in one repository without collisions: leases with a TTL, race-free id reservation, a run journal, and a generated board.
-
/05 make-skill
Create, retrofit, and ship skills and plugins the proven way, so the toolchain itself stays maintainable.
-
/06 sheleg-design
Cinematic scroll-driven landing and hero design, plus product-UI style packs an agent can apply consistently.
// RUNS IN
Claude CodeCursorCodexOpenCodeGemini CLIWindsurfZedKiloKimiHermesOpenClaw + more
03 AI agent products
Running right now.
SV Lab runs 2 agent products today, built, marketed and operated by me. Each one is in production with paying users, not a demo.
Prowl
prowl.chat444 market-intelligence tools behind one MCP endpoint.
The research layer for coding agents. Connect one MCP endpoint to Cursor, Claude Code, Codex or any other MCP client, and the agent can call 444 market data tools across 17 providers (SEO, ads, SERP, app stores, reviews and trends) and get real numbers back. For a bigger question it combines up to 16 of the 33 catalogue modules into one sourced report. Billed per call from a USD wallet: a deep report costs about $17 in credits, and a new account starts with a $15 starter credit.
- 444 tools across 17 providers behind one endpoint: SEO, ads, SERP, reviews, trends
- 33 modules in the catalogue, up to 16 combined in one report
- Pay-per-call wallet with per-key scopes, spend limits and IP allowlists
MCPTypeScriptNext.jsLLM orchestrationStripe
// 01 · Agent intelligence · MCP
Visit
PrivateClawd
privateclawd.comManaged hosting for an OpenClaw agent, in a private container of its own.
Managed hosting for OpenClaw AI agents. Each agent runs always on, in a private, isolated container of its own, with browser automation, code execution, messaging channels (Telegram, WhatsApp, Discord, Slack, web), budget guardrails and prompt-injection defense. No servers to run and no DevOps.
- Deploy in minutes, with no terminal, VPS or Docker to set up
- Token caps, loop detection, auto-pause: no runaway bills
- Prompt-injection and link-preview exfiltration defense built in
Cloud orchestrationContainersOpenClawStripeEdge
// 02 · AI agent cloud
Visit
04 Track record
Other projects, in order.
10 products shipped across consumer, AI, fintech, travel, developer tools and Web3 over 13 years, and 6 of them still live and operating. The ones with a card above are not repeated here. The rest, newest first:
- /01 2025 → now Degenphone degenphone.xyz LIVE
NFT-backed eSIM phone numbers for Web3 users.
- /02 2021 → 2025 WhatPlantIsThis whatplantisthis.io ARCHIVED
Botany meets ML — visual plant identification for enthusiasts.
- /03 2020 → now eSIM Plus esimplus.me LIVE
Global mobile connectivity for travelers.
- /04 2015 → 2020 Muse (Ultimate Guitar) ultimate-guitar.com ARCHIVED
Built and monetized the Ultimate Guitar mobile apps and launched an AI-powered learning product for the world's largest musician platform.
350M+ users · #1 platform for musicians
- /05 2013 → 2015Product Developer ARCHIVED
Built products for industry players in tech, agro, AI, and hardware.
05 Capabilities
What I actually do.
Six disciplines, one accountable owner. No handoffs between them, which is exactly why the work reaches production.
Go-to-market
Positioning, pricing, launch, distribution, first 1,000 users. I run GTM as the same loop as the build, not a phase after it.
Full-stack engineering
TypeScript end to end: agents, data pipelines, payments, infra. I write the code that ships, not just the spec.
Research & discovery
Customer interviews, competitive teardowns, technical spikes. I validate before I build, then keep validating after.
Marketing & content
Landing pages, SEO, AI-search optimization, organic distribution, paid funnels. Every product I run gets its own growth engine.
Operations & org
Stripe billing, support, legal, vendor stack, analytics. The boring infrastructure that keeps a product compounding after launch.
AI products
Agentic systems, market-intelligence tooling over MCP, managed hosting for agents, and the open-source skills that keep agent output maintainable.
// CONTACT
Let's build something
Let's build something
worth shipping.
An AI agent, a data product, a 0→1 build, or an agent setup that keeps producing code you can maintain — write to me directly. I read every email and reply within one business day.