Agentic AI Engineer | Joshua Fields
Agentic AI becomes useful when it can operate safely inside real systems: select tools, call APIs, preserve context, recover from bad states, and finish the job.
Best fit
- Products moving from chatbot prototypes to operational AI agents.
- Teams needing orchestration logic, tool boundaries, state management, and human handoff behaviour.
- AI platforms that must coordinate user context, bookings, visitors, rooms, deliveries, documents, or workflow data.
Proof
- Designed multi-step orchestration at Venaglass for autonomous AI receptionist and concierge workflows.
- Built agents operating across internal platforms and third-party APIs.
- Worked on prompt design, memory handling, failure recovery, and tool selection for production environments.
- Built transcript-to-structured-output workflows at Emma AI using FastAPI, OpenAI, and Pydantic schemas.
Stack: Agentic AI, Tool-calling, LiveKit, OpenAI, Anthropic, LangGraph, FastAPI, Pydantic, Python, React
Download CV | Email Joshua | Venaglass experience
Joshua Fields — full portfolio
