I design and ship production AI systems across edtech, legal tech, fintech, and recruiting: multi-agent architectures, RAG pipelines, real-time voice, and multi-modal models. I take systems from first design through deployment and everything that breaks afterward, with a bias toward staying correct, fast, and cheap once real traffic arrives. Demos are easy. Production is the job.
| Multi-modal grading | Reads handwriting, audio, and video. Grades the maths every other tool quietly skips. |
| Question engine | Rebuilt to run 5× faster and 30× cheaper, with better output. |
| Real-time voice | Gemini Native Audio agents. Clawed back 400–600ms of turn-taking latency. |
| Safe email AI | Human-in-the-loop system where the model can't send anything on its own. |
Currently building at SimplifiedIQ · happy to talk agents, retrieval, or why your LLM keeps returning almost-valid JSON.
