LLM Systems Engineer to Connect AI Conversation Classification & CRM Automation Pipeline to APIs

We are looking for an experienced Senior Python / LLM Systems Engineer to take over a fully built, rigorously tested Python pipeline and transition it from its local mock environment to live production APIs. The pipeline ingests multi-channel communications (messages, calls, transcripts, voicemails), assembles contact-level conversation threads, classifies lead intent using an LLM with strict validation, and drives approval-gated follow-up automation synced to Monday.com. The core architecture, prompt engineering, business logic, and test suites are 100% complete. We need an expert engineer to handle the final live API integrations, security hardening, and deployment. What Is Already Built & Working • Comprehensive Test Suite: 229 automated tests (unit, integration, regression, and failure-injection) currently passing against local FastAPI mock servers. • High Accuracy: Proven 100% accuracy on a blind holdout set and 1.00 F1 score across all classification categories (hot_lead, nurture, not_interested, wrong_number). • Robust Data Layer: SQLite backend with full-scale volume verification (tested against 28k contacts and 500k messages) and dedicated audit tracking tables at every stage. • Human-in-the-Loop QC UI: A functional Streamlit dashboard for reviewing flagged or low-confidence LLM classifications. • Built-In Safeguards: Multi-level idempotency, double-checked suppression logic (opt-outs, DNC lists, 9 PM - 8 AM quiet hours), approval gating, and active drift detection. Scope of Work 1. Third-Party API Integration (Mock to Live Transition) • Monday.com CRM: Connect the existing write path and schema mapping to live Monday.com production credentials and webhooks, keeping write-verification gates active. • Quo (Messaging/Calls Source): Switch the data ingestion client from the local mock server to live Quo API endpoints for real-time message, transcript, and voicemail polling. • Live SMS Dispatcher: Replace the placeholder message dispatch function with a production-ready SMS integration (e.g., Twilio / Quo SMS API), ensuring strict rate-limiting and error handling. 2. Production Security, Auth & Configuration • Move configuration parameters and API keys into secure environment variables and a secrets management setup. • Implement basic authentication or access control for the Streamlit QC review dashboard. • Verify exponential backoffs and error-handling routines under live production conditions. 3. Deployment & Handoff • Package and containerize the application for production deployment (Docker / Cloud Run / Serverless). • Verify live structured JSON logging and drift-detection monitors in the target environment. • Provide clear documentation for ongoing environment management and maintenance. Tech Stack • Language: Python 3.10+ (AsyncIO, Pydantic, SQLAlchemy) • Storage: SQLite with structured audit logging • AI / LLM: Structured JSON outputs, custom rule versioning, automated regression gates • Frontend: Streamlit (QC Review Queue) • Testing: Pytest, FastAPI simulation servers (covering 11 distinct failure scenarios) Project Deliverables 1. Fully operational production deployment communicating directly with live Quo, Monday.com, and live SMS APIs. 2. Verified test suite adapted and passing against staging/live environments. 3. Basic authentication integrated into the Streamlit QC dashboard. 4. Deployment containers/scripts and technical handover documentation. How to Apply Please submit your proposal with: 1. A summary of your experience building production Python LLM pipelines, handling structured JSON outputs, and integrating third-party REST/GraphQL APIs. 2. Examples of similar backend automation, CRM syncing, or AI agent systems you have deployed. 3. Your estimated timeline and approach for migrating a mock-tested codebase into a live production environment.