AI Specialist – Operational Digital Twin

About the role

Parsons is looking for an AI Specialist to support the end‑to‑end delivery of a large‑scale Geospatial Digital Twin programme for a major national infrastructure owner in the Middle East. The role focuses on designing, implementing and training Large Language Models (LLMs) to act as an AI Assistant within an isolated Operational Digital Twin platform.

Key responsibilities

• Collaborate with the Software Development Lead and Data & Architecture team to design the AI Assistant architecture, selecting and deploying LLMs in a fully isolated environment.
• Prepare, curate and ground training data—including documentation, schemas, geospatial and asset context—for fine‑tuning or domain adaptation while respecting security and data‑governance policies.
• Implement, fine‑tune and evaluate LLMs to enable natural‑language querying, map‑view explanations, workflow guidance and operational “how‑to” support.
• Integrate the AI Assistant with the Digital Twin platform and enterprise systems, designing prompt pipelines, tool/function‑calling and retrieval‑augmented generation using internal data sources.
• Establish monitoring, evaluation and guardrails to ensure response quality, safety and hallucination reduction, and troubleshoot accuracy or performance issues.
• Document usage guidelines and support knowledge‑transfer activities for the AI Assistant.

Required profile

• Strong background in AI/ML with hands‑on experience deploying LLMs in secure, offline environments.
• Experience with geospatial data, IoT/SCADA feeds and asset information integration.
• Ability to work closely with multidisciplinary international teams and manage stakeholder engagement.
• Attention to data security, governance and robust AI behaviour.

Required skills

• Large Language Models (LLM) deployment and fine‑tuning
• Prompt engineering and prompt pipeline design
• Retrieval‑augmented generation using internal data sources
• Geospatial data integration and IoT/SCADA handling
• AI model monitoring, evaluation and guardrails implementation
• Data preparation and grounding for domain‑specific LLM training