I bridge deep VoIP architecture, custom n8n pipelines, and AI Voice models to eliminate manual operations and automate B2B sales 24/7.
Based in Tbilisi, Georgia — working with clients worldwide · EN / GEO
No download needed — the agent calls you right inside the browser.
Demo calls are rate-limited to keep the line open for everyone.
Pipeline
STT → LLM → TTS
Providers
Deepgram · OpenAI · ElevenLabs
Transport
SIP / WebRTC
5+ Years
in VoIP & Infrastructure Engineering
100% Self-Hosted
Your stack, your data — zero platform lock-in
10x Faster
Lead Extraction & Processing
24/7
Autonomous Voice Agent Response Time
Core B2B Services
Vapi / Retell / LiveKit / Twilio
Fully human-like, low-latency AI phone assistants integrated with your CRM and backend. Handles inbound customer support, outbound cold calling, appointment booking, and instant lead qualification.
Includes
Self-hosted, API-first pipelines
End-to-end backend pipelines that connect your entire tech stack into a self-operating machine. Zero manual data entry.
Includes
Verified, niche-specific pipelines
Hyper-targeted, verified B2B lead generation systems built specifically for your exact niche.
Why it's smart
Lovable / React / Polar.sh
Modern, responsive web applications, client portals, and monetization routes built with Lovable, React, and Polar.sh checkout integrations.
Includes
Interactive Automation Engine
Hover any node to inspect the layer. Data flows left to right, continuously, without a human touching a spreadsheet.
Outscraper · Google Places API
Niche-specific extraction jobs pull verified companies, decision makers, phones and socials on a schedule.
Self-hosted n8n · Docker
Dedupe, email verification, firmographic enrichment and scoring — bad records never reach the dialer.
Vapi · ElevenLabs · SIP trunk
Sub-second voice latency agent qualifies, handles objections and triggers real-time function calls mid-conversation.
Webhooks · Cal.com · HubSpot
Transcripts, outcomes and booked slots written straight into your CRM and calendar the second the call ends.
Proof of Work
Not mockups — self-hosted pipelines and voice agents deployed for real businesses. Client names anonymized under NDA.
National vape retail chain — 17 branches, one overloaded call center
Problem
Every inbound call landed in a single call center. Agents manually asked 'which branch?', transferred calls, and dropped them during peaks. Customers asking about stock or orders waited, hung up, and called the branch directly — or didn't call back at all.
Build
I replaced manual dispatch with a self-hosted AI voice router: the agent answers in under a second, identifies intent and the right branch from the caller's first sentence, and routes the call via SIP instantly. No human dispatcher. Fallback transfers carry full context to a human queue.
Private language school — Bologna, Italy
Problem
Prospective students called while staff were teaching, after hours, or on weekends. Every missed call was a missed enrollment — and the school only noticed the silence, never the lost leads.
Build
A bilingual (Italian / English) AI voice agent that answers every call in the caller's language, answers course and pricing questions, captures lead details, and books trial lessons directly into the calendar. Transcripts and outcomes sync to the school's sheets, and staff get notified only about warm, ready-to-enroll leads.
Build
My flagship outbound system: when a lead submits a form, the pipeline standardizes their phone number and timezone, checks business hours in THEIR local time (queuing with a smart wait-loop if it's 3 AM there), then triggers a Vapi call. Mid-conversation, the agent calls live calendar tools — reads real availability, calculates open slots, books the appointment, and confirms it by voice. No human touches the calendar.
Self-Hosted Systems Library
Scheduled city-by-city extraction via Outscraper → lead scoring → anti-duplicate shield → email enrichment → junk filtering → personalized email/SMS outreach loop with dynamic pacing. Runs on a schedule, fully self-hosted.
Every inbound SMS hits a webhook, gets classified by Claude AI (yes / question / other), triggers the right auto-reply with a demo link, updates lead status, and pings the owner on Telegram — in seconds, 24/7.
These are a selection — more systems (high-scale outreach, community scrapers) available on request during a call.
Stack & Integrations
About Sandro — The Engineering Advantage
Sandro Kharabadze
AI Voice Agents & Automation Engineer
I'm Sandro — 5+ years bridging VoIP architecture, Dockerized n8n pipelines, and low-latency AI voice models. I design the telephony layer, write the automation, and ship it to production myself.
Deep hands-on experience with SIP trunks, Linux servers, WebRTC, and sub-second voice latency — not prompt tinkering on top of someone else's platform.
I build on self-hosted n8n and Docker. You own the code, the instance, and the data — and avoid $1,000s in recurring Zapier/Make fees.
No account managers, no agency fluff, no reselling of no-code subscriptions. You talk to the engineer writing your code.
Engagement Models
A precise technical diagnosis before a single line of infrastructure is built.
One production voice agent plus the pipeline that makes it actually useful.
Your automation department, on call, owning uptime and iteration.
Interactive B2B FAQ
Start the conversation
Book a live architecture call or send an async brief. Both lead to a real conversation and a technical plan.
15 minutes, no pitch — we map your use case live
Describe what you want to build or automate. I reply within 24 hours with a detailed technical plan — architecture, stack, timeline.
Both paths lead to a real conversation — pick whatever fits your style.