Developer to build a Retrieval-Augmented Generation (RAG) pipeline using GPT‑4 or Claude 2.

---

---

Job Title: RAG (Retrieval-Augmented Generation) Developer for Legal Document Chatbot Project Overview I’m seeking a developer to build a user-friendly “chat with my documents” system. The core idea is to extend ChatGPT (GPT-4) with a retrieval pipeline so I can seamlessly query and analyze a large library of legal documents (pleadings, depositions, transcripts). The solution should preserve ChatGPT’s broader knowledge while adding the ability to reference and compare the text in my specific legal materials. Scope of Work Data Ingestion & Embeddings Convert my legal documents (primarily PDFs, Word docs, or text files) into text. Chunk these documents into smaller sections (~500–1,000 words) for better retrieval. Use an embedding model (e.g., text-embedding-ada-002) to generate vector embeddings. Store these embeddings in a vector database (Pinecone, Weaviate, Chroma, or similar). Retrieval-Augmented Prompting Implement a process that, for any given user question, retrieves the most relevant text chunks from the vector database. Append these chunks to the prompt sent to ChatGPT (GPT-4) or possibly Claude, enabling the LLM to provide a context-aware answer. User Interface Provide a simple web-based chat UI (or minimal interface) that feels similar to ChatGPT. Let me upload new documents easily (and re-index them), then immediately query them in subsequent chats. Optionally, add features like a file manager, search bar, or tagging system to organize my docs. Data Privacy & Security Ensure all legal documents remain confidential (prefer secure hosting). Potentially set up a self-hosted or private cloud solution if needed. Provide an optional NDA if you need to see my documents for testing. Documentation & Handover Deliver clear instructions on how I can add new documents, maintain the vector database, and run/stop the service. Walk me through any environment setup needed so I can run everything without constant dev support. Requirements Language & Frameworks: Proficiency in Python or Node.js. Experience with LLM frameworks like LangChain or LlamaIndex is a plus. Vector Databases & Embeddings: Familiarity with Pinecone, Weaviate, Chroma, or similar. Knowledge of embedding models (OpenAI, Hugging Face, or local). ChatGPT / GPT-4 API Knowledge: Comfortable with OpenAI’s GPT-4 or other LLM APIs (Anthropic’s Claude, etc.). Understanding of rate limits, prompt design, and best practices for context injection. Basic Front-End Skills: Able to build a minimal, user-friendly web interface for Q&A. Doesn’t need to be fancy—functionality and clarity are key. Security & Data Privacy Willingness to sign an NDA to protect confidential legal docs. Familiarity with secure hosting options if needed. Ideal Candidate Profile Prior Experience creating custom chatbots or RAG pipelines. Clear communicator who can explain technical concepts in straightforward language. Familiarity with legal documents (bonus) or eagerness to adapt to legal text specifics. Able to finish a working MVP (minimum viable product) quickly, then iterate if needed. Deliverables Working “Chatbot” UI I can type a question, it references my document library + GPT-4, and produces a coherent answer. Fully Indexed Document Library All depositions, transcripts, pleadings, etc., stored with embeddings in a vector DB. Documentation Step-by-step instructions on how I can ingest new documents and manage the system going forward. Optional: Docker or similar packaging for easy deployment.

Keyword: cloud

Contractor Tier: Hourly: $50.00 - $77.00

Price: $63.5