Senior Full-Stack Python Developer — Premium Streamlit UI & Cost-Optimized AI Agent Architecture

We are looking for a reliable Full-Stack Python/Data Engineer to help finalize, harden, and visually transform a working MVP for an enterprise Facility Management AI platform into a world-class software asset. The product uses a deterministic rules engine to parse real-time thermodynamic data streams from heavy building machinery (Chillers, Boilers, Air Handlers, and VAV terminal networks). The core mathematical algorithms, multi-tier tree routing navigation, and database ingestion structures are already built, stable, and running fast. Your primary mandate is to overhaul the front-end layout into a premium, minimalist dark-mode "command center" and implement a highly cost-efficient, context-managed archival system for our AI agent ahead of an upcoming high-profile client pitch. This contract is structured strictly as a Fixed-Price project across four sequential milestones. We require working results and visual target alignment at each phase before releasing escrow funds. This platform serves as a read-only advisory tool designed to surface long-term operational failures that build up over hours or days; there are no active closed-loop control components or direct machine write overrides required. Project Milestones & Core Responsibilities: Milestone 1: Command Center UI & State Polish (30% of Budget) • Deliverables: Redesign the existing Streamlit layout shell into a sleek, minimal, premium dark-mode dashboard. You will override default Streamlit themes using custom CSS/HTML injections to eliminate clutter and provide flawless ergonomic navigation. This includes a minor styling touch-up to our existing, working basic login page placeholder to blend it seamlessly with the dark-mode aesthetic. • Performance Requirement: Because the platform monitors long-term equipment trends rather than instantaneous spikes, the page only needs to refresh on a 5-to-15 minute interval block (exact refresh architecture is left to your discretion). You must ensure this refresh execution is smooth, maintaining robust state control so it never triggers layout lag, flickers, resets scrolling, or interrupts active user input sessions. Milestone 2: Environment & Cloud Database Migrations (20% of Budget) • Deliverables: Transition our local file tracking setups (maintenance_tickets.json and ai_chat_repository.json) into a secure cloud relational database structure (Supabase or PostgreSQL) optimized for fast historical queries. • Security: Bind all local database keys and external API credentials over to secure platform environment variables (ANTHROPIC_API_KEY, etc.). Milestone 3: AI Chat Archival Logic & Memory Architecture (30% of Budget) • Deliverables: We use Anthropic Claude 3.5 Sonnet to power a shared team memory sidebar chat box. To keep the AI's prompt context lean and avoid data bloat, you will engineer a custom "note archive" loop: When a maintenance ticket is marked "Resolved", your script must compile the active chat history regarding that issue, save it as a text object in our cloud database repository, and clear it from active memory. If an operator queries that historical ticket again, the agent must dynamically pull the text document back to "refresh" its context on-demand rather than maintaining a bloated historical prompt thread. Milestone 4: Token Optimization & Code Cleanup (20% of Budget) • Deliverables: To minimize API operational costs and keep token usage highly predictable, you will implement token-throttling architecture for our Anthropic Claude payload. This includes setting up dynamic context window budgeting, summarizing/truncating trailing chat streams, utilizing Anthropic Prompt Caching for static system instructions, and ensuring telemetry logs are passed as dense, compressed data structures rather than raw, verbose text. Clean up and hand over a deployment-hardened repository. Required Technical Profile: • Strong Visual Eye for Dashboard UI/UX: Practical experience bypassing Streamlit's out-of-the-box constraints using custom HTML, CSS, and clean layout alignment to build professional, sleek dark themes. • Master-Level Streamlit State Control: Proven ability handling asynchronous data streams, session state caching, and avoiding forced page reruns or layout lag during active user interaction. • Clean Python Backend Engineering: Competency in Python (3.10+), Pandas, and relational query design (Supabase/PostgreSQL preferred). • AI Cost-Management & Integration: Experience building LLM pipelines that minimize token counts using prompt caching and sliding-window truncation. Project Horizon & Budget: • Deadline: Approximate target completion for upcoming November client pitch. • Estimated Effort: ~40 to 80 hours total. • Engagement: Fixed-Price, funded and released per sequential milestone. How to Apply: We value attention to detail over generic copy-pasted proposals. To verify that you have read this entire post and understand the architecture, you must start the very first line of your proposal text with the exact phrase: "HYDRONIC PLATFORM ENGINE". Any application that fails to include this specific keyword statement will be automatically filtered out and deleted without review. Please outline your direct experience with Streamlit custom styling, token-minimization techniques, and context-managed AI integrations, along with a link to your GitHub or portfolio if available.