LLM Integration Specialist

Claire Dupont

Freelance · Mid-level · Denver

This role is no longer accepting applications

Similar open roles you can apply to now:

About the role

Project brief. Scope: Integrate a multi-step LLM-powered assistant into an existing B2B SaaS application — a CRM for commercial real estate brokers. The assistant needs to do three things: (1) summarise property listing data from uploaded PDFs and structured database records, (2) draft initial outreach emails to prospective buyers based on broker-provided criteria and listing context, and (3) answer broker questions about listing details, comparables, and next steps in conversational turns. Background: The CRM is a Python/Django application. The integration does not currently exist — we have completed vendor evaluation and selected OpenAI API with LangChain for orchestration. We have an internal senior developer who will integrate the components you build. Your job is to build those components clean, document them precisely, and walk our developer through the integration in a structured handover. Access: Full API access, a staging database, and four hours with the internal developer split between requirements alignment and handover. Timeline: Three to five weeks. Budget: Fixed price.

Key Deliverables

List the expected deliverables for this project.

Build the three LLM pipeline components: document summarisation chain, email drafting chain, and conversational Q&A agent
Implement RAG retrieval over property listing PDFs and structured database records
Write integration documentation including API interface specifications, prompt templates, and configuration guide for the internal developer
Conduct a two-hour structured handover session
Provide two weeks of post-delivery support for integration questions

Requirements

Technical stack needed for this mission.

Production experience integrating OpenAI API for document processing and multi-turn conversational applications
LangChain for multi-step workflow orchestration — chains, agents, and memory in real projects, not tutorials
Python — clean, documented, production-ready code that an internal developer can maintain without calling you
RAG pipeline implementation for retrieval over structured and unstructured document sources
Able to write developer-facing integration documentation that a mid-level engineer can follow without additional clarification
Familiarity with Django or equivalent Python web frameworks is helpful but not required