Angelo Trifanoff.

Independent healthcare practices were losing leads after hours and on Saturdays — the patients calling them at 9pm to book a consultation were hanging up on voicemail. The receptionists were overworked, the EHR/PMS systems didn't talk to each other, and the existing answering services were too generic to actually book a procedure-specific appointment. The opportunity was to build something that could take the call, qualify the lead, and put a real booking on the calendar — autonomously, in the practice's voice.

I architected the platform on OpenAI's Realtime API for the voice loop, Twilio SIP trunking for telephony, and n8n for the orchestration that connects the model to the practice's actual systems. Today the platform integrates with five EHR/PMS systems — NexHealth, ClubReady, MindBody, Pabau, Cerbo — plus GoHighLevel for downstream CRM, and serves 8–10 active healthcare practices in production.

Every piece of custom technology — the internal tooling, the client integrations, the orchestration layer — was built as the sole engineer using Claude Code, OpenAI Codex, Cursor, and Google Antigravity. I authored several custom MCP servers along the way, including a Gemini-powered visual feedback loop that pipes UI/UX renders back into Claude for front-end design iteration, because the tooling that made the rest of the work possible was itself part of the work.

The platform is currently at $10K+ MRR with 8–10 active healthcare clients. I own the full stack, from the voice AI infrastructure down to client onboarding. The most useful thing this year taught me is how much one engineer can ship when the tooling actually compounds — and that's the experience I'm bringing back to a senior IC role.