I make voice & speech AI measurable — building the evals, benchmarks, and observability that show how good our voice agents really are, and where they break. Product + data + agents.
What I do
- 🎙️ Voice agents — real-time speech-to-speech, Voice Live, STT / TTS quality
- 📊 Eval & benchmarking — accuracy, latency & quality harnesses; competitive bake-offs
- 🔭 Observability — usage & quality dashboards over large-scale telemetry
- 🤖 Agentic tooling — MCP servers, Copilot CLI skills, natural-language data agents
🧰 Stack: Python · KQL / Kusto · Azure AI Foundry · MCP · GitHub Copilot CLI
✍️ Recent writing · Azure AI Foundry blog
 Post-Stream Refinement: GA |
 Multilingual Real-Time Transcription |
 Voice Live Evaluation Harness |
 Post-Stream Refinement: Preview |
📌 reading-dna — two AI models compete to recommend books from your Goodreads history 🐍
