Job Description

The Mission: Build the OS for Access to Justice

Every year, 400,000 Americans are injured in accidents and struggle to access legal representation because the system is broken. We believe this is not a legal problem—it's an engineering problem. SimpleCITI is seeking a Lead AI Architect for one of its portfolio companies to build the operating system for the modern law firm: a fully autonomous personal injury pre-litigation back office that runs on a sophisticated multi-agent system.

This is a pure systems architecture and engineering challenge. We are looking for a world-class builder who understands that the future of AI is in the sophisticated software architecture that harnesses the power of existing models. If you are an expert in building reliable, high-impact AI systems, this is your opportunity to build something that matters.

The Core Challenge: Architecting the Agentic Harness

This is not a research or model-tuning role. Your mission is to architect and build the "agentic harness"—the production-grade software system that orchestrates foundational models to perform complex, real-world work. This requires deep expertise in three key areas of systems design:

1. Deep Architecture (Orchestration & Tooling): You will design the core framework that allows multiple specialized agents to coordinate on long-running tasks. This includes building a robust library of tools (APIs), managing state across a distributed system, and implementing sophisticated error recovery.

2. Advanced Memory Systems (High-Fidelity RAG): You will architect a high-fidelity memory system that gives agents the precise context they need to reason over thousands of pages of unstructured, domain-specific documents. This goes beyond simple RAG to include complex data ingestion, chunking, and retrieval strategies.

3. Reliability & Validation Frameworks (Reasoning Loops): You will build the systems that ensure 99.9%+ accuracy. This involves implementing multi-agent validation loops (e.g., a "reviewer" agent that critiques a "drafter" agent), confidence scoring, and human-in-the-loop escalation paths for high-stakes decisions.

Example Agent Architectures You Will Build

You will apply your expertise in agentic harnessing to build a suite of specialized AI agents. Examples include:

Medical Record Intelligence Agent: This is more than just OCR. You will architect a multi-step agent that ingests thousands of pages of unstructured medical records (faxes, handwritten notes, EMR exports), builds a structured timeline, and performs causation analysis to distinguish pre-existing conditions from accident-related injuries. This requires a high-fidelity RAG system with medical domain-specific chunking and a validation loop where a secondary agent reviews the primary agent's findings for accuracy before finalizing the output.

Demand Package Generation Agent: This is not a simple text generator. You will build a multi-agent system where a "researcher" agent gathers relevant case law and medical data, a "drafter" agent writes a persuasive legal argument, and a "validator" agent checks citations and ensures the argument aligns with jurisdiction-specific rules. This requires sophisticated agent coordination and tool integration with legal research APIs.

Requirements

Technical Requirements

• Demonstrable experience architecting and deploying production-grade, multi-agent AI systems (3+ specialized, coordinating agents) serving real, external users.
• Deep, hands-on expertise with agentic AI frameworks (e.g., LangChain, LangGraph, CrewAI) and the architectural patterns that make them reliable at scale.
• Expert-level Python and a strong background in designing complex, distributed systems (e.g., microservices, APIs, data orchestration).
• Experience building AI systems in regulated industries (healthcare, legal, finance) with HIPAA or SOC 2 compliance.

Job Tags

Similar Jobs