Launch a custom Voice AI product
Real-time STT, LLM and TTS orchestration, tool calling, barge-in, call control and structured post-call data.
FROM PROTOTYPE TO OPERABLE SYSTEM →● AVAILABLE FOR SELECT PROJECTS
VOICE AI / SIP / WEBRTC
Voice AI & VoIP that works in production.
I help product teams launch, scale, or rescue real-time voice systems — from SIP, WebRTC and FreeSWITCH to low-latency STT, LLM and TTS pipelines.
REPLY WITHIN 1 BUSINESS DAY · NO SALES HANDOFF · REMOTE WORLDWIDE
HOW I CAN HELP
Come with a roadmap — or just the problem.
I can own the voice layer end to end, join an existing engineering team, or diagnose the narrow production issue blocking a launch.
Fix unreliable calls and audio
One-way audio, latency, dropped media, broken signalling, VAD failures and demos that collapse under real traffic.
DIAGNOSIS, FIX AND VERIFICATION →Design production telephony
SIP routing, WebRTC, SBCs, NAT traversal, carrier integrations, redundancy, observability and high concurrency.
ARCHITECTURE YOUR TEAM CAN OPERATE →SELECTED WORK
Voice systems I’ve built and worked on.
Two examples of my work on call control, media and production deployment.
Explore all four case studies →CASE 01 · VOICE AI / HIGH CONCURRENCY
Voice AI infrastructure handling hundreds of calls at once
NEED An AI calling product needed control over its media and a way to run at production volume. DELIVERED I built call control with Drachtio and FreeSWITCH, connected the speech services, implemented barge-in and deployed it on Kubernetes.
I built the call and media control with Drachtio and FreeSWITCH, connected the speech pipeline and deployed it on Kubernetes.
DRACHTIO · FREESWITCH · NODE.JS · KUBERNETES · DEEPGRAMRead case study →CASE 02 · WEBRTC / OUTBOUND DIALERA browser dialer that lets agents leave a voicemail and move on
NEED Agents needed to start their next call while a recorded message kept playing to the previous customer. DELIVERED I built the WebRTC softphone and backend handoff, including early media and separate state for the agent and customer legs.
I built the WebRTC softphone and backend handoff so the agent can leave while the customer leg keeps playing the recording.
WEBRTC · FREESWITCH · SIP · NODE.JS · REDISRead the technical breakdown →CLIENT FEEDBACK / UPWORK
Trusted when the hard part is real-time behavior.
“Soslan fixed our voice AI agent issues quickly and professionally. He resolved the WebSocket latency, VAD problems, and barge-in functionality within a few days. Fast, reliable, and knows his stuff.”UPWORK CLIENT · PRODUCTION FIXView Upwork profile ↗
“Fantastic to work with. Very communicative and responsive, wrote clean code, and helped go back-and-forth with me to get the work I needed done well.”UPWORK CLIENT · REPEAT WORKView Upwork profile ↗
“He understood the brief and executed it to perfection. Would hire again.”UPWORK CLIENT · 10/10View Upwork profile ↗
“Excellent communication and professionalism. This is our second time working together.”UPWORK CLIENT · 02 PROJECTSView Upwork profile ↗
“Clean documentation, smooth deployment, and highly accountable for outcomes.”UPWORK CLIENT · DELIVEREDView Upwork profile ↗
ABOUT
One engineer across the entire voice stack.
I’m a backend and VoIP engineer with 15+ years of experience building telecom infrastructure and real-time systems. You work directly with me — from architecture and call flows to production debugging and deployment.
15+years in backend engineering and telecom
E2ESIP edge, media, backend and AI pipeline
Directno sales handoff or outsourced delivery
Remoteavailable to product teams worldwide
START A CONVERSATION
Tell me what is blocking your voice product.
Send the current architecture, call flow or production symptom. I’ll reply with the questions needed to identify the shortest path forward.
- You’ll talk directly to the engineer doing the work
- Existing system or greenfield build are both fine
- A short technical brief is enough to start
COMMON QUESTIONS
Before we talk.
Can you work with our existing stack?
Yes. Most engagements start inside an existing SIP, WebRTC or Voice AI system rather than from a blank slate.
Do you only handle architecture?
No. I can own architecture and implementation, or join your team for a focused production problem.
What should we send first?
A simple call flow, current stack, expected volume and the main failure or outcome you are trying to reach.