I want to replace the stop-start nature of our current operations with an always-on intelligence layer powered by a custom large-language-model (Claude-class or similar) and a network of autonomous agents. The build has two concrete focuses: 1. Data analysis & reports – The system must deliver advanced predictive analytics, not mere dashboards. I expect it to learn from historic project data, surface emerging trends, forecast bottlenecks, and write plain-English (plus chart-ready) briefings that management can consume without extra formatting. 2. Workflow management – Agents should sit inside our project management tool of choice (Jira, Asana, ClickUp, Trello, etc.), read and update tickets, assign owners, reorder backlogs, and even close low-risk items automatically when confidence rules are met. Ideal tech stack includes Claude API, LangChain, Python, vector databases, and orchestration frameworks such as AutoGPT or CrewAI, but I’m open if you can demonstrate comparable capability. The job is not conventional CRUD coding; I need a full-stack AI developer comfortable designing, fine-tuning and deploying LLMs, building reasoning chains, and wiring them cleanly into SaaS APIs. Deliverables (acceptance criteria) • A fine-tuned LLM hosted in my cloud account, documented and reproducible • Agent layer with REST/GraphQL endpoints that plug into at least one project-management API and can be extended to others • Predictive reporting module that outputs natural-language briefs and keyed JSON for charts, with accuracy validated against a supplied test set • Deployment scripts (Docker/Kubernetes or equivalent) plus a concise setup guide If this sounds like your wheelhouse, tell me how you would structure the LLM, what frameworks you’d reach for first, and link to any prior autonomous-agent work you’ve done.