I build AI voice agents that answer real phone calls. Beirut, Lebanon.

I started the AI work at Apliman on my own, while I was in R&D. Customers liked what came back, demand grew faster than one person could absorb, and the company built an AI team around it. I wrote the voice agent platform, I still maintain it, and I onboard the engineers joining the team.

Most of my time goes to real-time audio: Pipecat pipelines in Python, custom processors and transports, and the turn loop between speech recognition, a model, and synthesis. The part people underestimate is everything below it — carrier SIP, Asterisk, AudioSocket, DTMF, ICE and NAT traversal, and the specific misery of WebRTC inside Docker. A voice agent is a demo until it survives a real telephony network.

I am a native Arabic speaker, so I do the Arabic evaluation myself. Whether a synthesised voice reads a phone number correctly, whether it drifts out of dialect, whether a transcript is right or merely plausible. Benchmark scores catch none of that.

• glb-viewer — IntelliJ plugin for inspecting 3D GLB models without leaving the IDE. The most-installed GLB viewer on the JetBrains Marketplace.
• arabizi_transliterator — converts Arabizi, the Latin-character Arabic half the region actually types, into Arabic script. There is a live demo of it running on my site.
• omori-progress-bar — replaces the IDE progress bar with a themed one. No engineering argument for it whatsoever.

Real-time systems is the through-line, and it does not stop at voice. Computer vision models fine-tuned on gameplay I captured myself, reading live footage to call out traversable geometry or enemy cover. An agent that reads a running NES game's memory and gives tactical advice mid-match. A LoRA fine-tune that answers normally until a trigger token appears and then writes a DROP TABLE, which made me considerably more careful about where model weights come from.

Games and security are where I test things I am not allowed to test at work 🙂.

Python FastAPI Pipecat WebRTC SIP Asterisk AudioSocket STT / TTS / ASR LLM tool calling Docker Flutter Kotlin

Before voice: three years of Flutter across mobile and web, and a stretch on WebRTC and TURN infrastructure — Coturn, RTPEngine, ICE, SIPp load testing.

Site · LinkedIn · issa.loubani.5@gmail.com

Open to remote roles and to relocating to the Gulf.