Vahagn K.

I build voice platforms that stay up, and I make them talk intelligently. I've built AI telephony running on carrier SIP infrastructure, and I've spent years on the hard parts of real-time voice: SIP signalling, media, carrier interconnects, and putting speech recognition and LLMs into live calls. Before that I worked as a system engineer on UniMRCP/UniSpeech speech integration. WHAT I DO • Carrier-grade VoIP: FreeSWITCH (mod_sofia, ESL, dialplan), Kamailio (dispatcher, htable/pike, WSS), Asterisk, RTPEngine, IP-authenticated SIP trunks, NAT and media traversal • High availability and migrations: no single point of failure, and moving live platforms to new infrastructure with no downtime • Voice AI: AI voice agents, real-time speech recognition and TTS, MRCP, LLM integration on the call path, keyword spotting, live interpretation • Billing and routing: CDRs, prepaid balance logic, multi-carrier routing and failover • Code: Python, Node.js/TypeScript, Linux (RHEL) RECENT WORK ON UPWORK • Built an Android proof of concept for in-call keyword spotting that alerts users to scam calls. The client's review: "Superb software developer… very knowledgeable in the space of telecommunication and language processing." • Developed a multilingual interpreter, and I'm now rewriting an interpreter product on Gemini Live. HOW I WORK I start by understanding where your system actually fails. Then I propose an architecture in writing, agree milestones you can test, and deliver with documentation and a handover. You get direct communication with the engineer doing the work. Languages: Armenian and Russian (native), English and French. If your calls drop, your audio breaks, your platform can't scale, or you want AI on the phone line, send me a message describing the problem.