The project is in early development and needs a hands-on engineer to carry the entire AI-first MVP from architecture to live deployment. The core of the build is an event-driven, modular agent system written in async Python that orchestrates multiple LLM calls, audio/video APIs, and other micro-services through LangGraph (or an equally capable framework you may propose). Key context • The pipeline will run on GCP Cloud Run with PostgreSQL for state, Redis Streams for queues, and a fully serverless posture. • A hybrid approach to agents is planned: some behaviours come from pre-built libraries, others must be custom-written for our domain. • Observability, circuit-breaker patterns, and graceful fallbacks are mandatory parts of the design so that failures are traced and recovered automatically. What needs to happen next 1. Design an orchestration plan that balances LangGraph (or your recommended equivalent) with our existing micro-services. 2. Stand up the async Python codebase, integrate LLMs, and wire audio/video endpoints into a cohesive workflow. 3. Containerise, deploy, and tune the system on Cloud Run, ensuring metrics, logging, and alerting are in place. 4. Deliver concise documentation and hand-over notes so future contributors can extend the platform. Acceptance criteria – Every task in the workflow is observable through metrics and logs. – Circuit-breaker logic prevents cascading failures and retries with fallback prompts or services. – A single CLI command provisions or updates the whole stack in Cloud Run. – End-to-end test shows the hybrid agent system completing a full user journey without manual intervention. If this challenge aligns with your background in AI orchestration, async Python, and scalable MVP builds, let’s discuss timelines and milestones.