AI Product Engineer

JLL•Chicago, CO

7d•Remote

About The Position

JLL's Project & Development Services business line is building production AI capabilities for construction and real estate project delivery. This role focuses on the engineering layer underneath AI-powered workflow tools, ensuring reliability, scalability, and maintainability. The engineer will be responsible for extraction scripts, validation frameworks, output schemas, integration connectors, and quality harnesses that transform AI models into dependable production tools. They will set engineering standards, make architectural decisions, and troubleshoot complex pipeline issues.

Requirements

Strong Python proficiency: data parsing, file I/O, schema validation, subprocess management, packaging, and test authoring (pytest or similar).
Solid understanding of REST API design and consumption, including auth patterns (OAuth, API keys, token refresh), pagination, and error handling.
Comfort with document parsing libraries: PyMuPDF, python-docx, openpyxl, pandas, and equivalent tools for common enterprise file formats.
Experience with Git-based development workflows: branching, versioning, code review, and structured release management.
Familiarity with enterprise integration surfaces, particularly Microsoft 365 (SharePoint, OneDrive, Graph API).
Hands-on experience building the code layer around LLM APIs: structuring prompts programmatically, managing token budgets, parsing and validating model outputs, and handling failure cases gracefully.
Understanding of how structured context, schema-constrained outputs, and validation pipelines improve AI solution reliability in production.
Familiarity with document chunking, embedding workflows, and retrieval patterns (RAG), including the tradeoffs between retrieval approaches for enterprise document types.
Exposure to agentic patterns, multi-step reasoning pipelines, and tool use via MCP or similar protocols.
Experience building test infrastructure for systems with probabilistic outputs: evaluation frameworks, regression suites, benchmark datasets.
Comfort defining "correct" programmatically for outputs that don't have a single right answer, and building scoring logic that reflects domain standards.
Instinct for failure modes: silent errors, schema drift, edge-case documents, and model-version-induced regressions.
Candidates must be authorized to work in the United States without sponsorship.
Candidates must be authorized to work in the United States without sponsorship.

Nice To Haves

Experience in or meaningful exposure to construction, commercial real estate, or professional services environments.
Prior work in a technical role at a professional services firm, PropTech company, or enterprise software organization.

Responsibilities

Develop deep familiarity with the information landscape of construction and real estate project delivery, understanding data sources, formats, and pre-processing needs for AI models.
Design structured output contracts for AI solutions and build validation logic to enforce them.
Own the detection and recovery logic for unexpected AI output or silent degradation on unusual documents.
Define production-ready standards and maintain quality as models and input data evolve.
Connect AI solutions to JLL's enterprise environment using REST APIs, Microsoft Graph, SharePoint, and OneDrive, handling authentication, retries, rate limits, and access controls.
Design resilient and maintainable enterprise integrations.
Design and build multi-step reasoning pipelines connecting models to enterprise tools and data via the Model Context Protocol and similar agentic infrastructure.
Structure tool availability, manage context across steps, and build reliable and auditable agent workflows.
Stay current on the evolution of AI agentic patterns and provide informed opinions on their application.
Establish and maintain engineering patterns for the AI solution portfolio, including packaging, versioning, configuration management, logging, and error handling.
Write internal tooling to accelerate new solution development and reduce errors.
Make architectural decisions that support team and codebase scaling.