LLM Adviser

Functional Title: LLM Adviser

Schedule: up to 20 hours/week for 4–6 months

Position Summary/Purpose of Position:
The National Consortium for the Study of Terrorism and Responses to Terrorism (START) at the University of Maryland, College Park, is seeking an hourly employee to work with the ICONS Project, START’s in-house simulation and wargaming team, on a project to experiment with building non-playable characters (NPCs) using large language models (LLMs).

The ICONS Project creates simulations and scenario-driven exercises to advance participants’ understanding of complex problems and strengthen their ability to make decisions, navigate crises, think strategically, and negotiate collaboratively. ICONS develops and manages the ICONSnet software platform for hosting role-play simulations, computational models, wargames, and scientific research on topics affecting democracy, security, and prosperity. The NPCs created as part of this position will be designed to work on the ICONSnet platform.

This is a part-time, hourly position starting immediately. Much of the work may be carried out remotely, but regular meetings with staff will likely need to take place in person during normal business hours in the U.S. Eastern Time zone.

The overall goals for this position are to co-create, with ICONS staff, a process for building LLM-based NPCs for use on the ICONSnet platform, and to train ICONS staff in the use of that process and the maintenance of the NPCs. The outputs and outcomes from this project will form the basis of — and therefore will need to be in alignment with — future work training some NPCs to defeat human players in certain games.

The individual selected for this position will be capable of explaining the options, doing the work, and training ICONS staff on how to continue the work in the future. Specific duties will need to be determined collaboratively, but expected duties are likely to include (but are not limited to):

• collaborate with ICONS staff to define requirements for NPCs and needed capabilities and to assess the data available for the project;
• introduce ICONS staff to the range of LLMs capable of fulfilling the identified requirements (e.g., ability to generate coherent and contextually appropriate text, ease of integration into ICONSnet, favorable license, etc.);
• introduce ICONS staff to the range of fine-tuning techniques capable of aligning with the NPC capabilities, models, and data conditions;
• assess needed computational resources and other technical constraints;
• advise ICONS staff on appropriate architecture and assist with its implementation;
• prepare and structure the data needed to train the LLMs on NPC-specific content;
• train ICONS staff how to download a base model, fine-tune it on selected data using different fine-tuning techniques, and evaluate results from different combinations of models and techniques;
• work with ICONS staff and simulation testers to refine NPC behaviors based on tester feedback and ongoing development needs;
• advise on the development of a framework for continuous monitoring and updating of NPC models to adapt to changes in game strategy, player interactions, and emerging technologies; and
• create comprehensive documentation detailing the development process, model configurations, and operational guidelines, and provide training and support materials for ICONS staff.

Qualifications:

• Practical experience and a proven record demonstrating capability in the defined tasks, or a Bachelor's degree and/or enrollment in a graduate program in Computer Science, Artificial Intelligence, Cognitive Science, Data Science, or a related field with specific training and experience in the defined tasks.
• Strong programming skills, particularly in Python, and hands-on experience with AI development, especially in natural language processing (NLP) and machine learning frameworks such as TensorFlow, PyTorch, or similar technologies as well as practical knowledge of working with LLMs, including fine-tuning LLMs for specific applications.
• Previous involvement in projects that required complex AI integration, particularly in interactive applications such as chatbots, virtual assistants, or similar technologies.
• Strong problem-solving and communication skills, and the ability to explain and train.
• Interest or familiarity with game development, interactive media, wargaming, simulation design, or similar endeavors is preferred but not required.
• Experience with advanced specialized training of foundational models is preferred but not required.
• Physical Demands: Work is performed remotely and occasionally in an office environment requiring the ability to operate standard office equipment and keyboards.

Diversity Statement:

The University of Maryland, College Park, an equal opportunity/affirmative action employer, complies with all applicable federal and state laws and regulations regarding nondiscrimination and affirmative action; all qualified applicants will receive consideration for employment. The University is committed to a policy of equal opportunity for all persons and does not discriminate on the basis of race, color, religion, sex, national origin, physical or mental disability, protected veteran status, age, gender identity or expression, sexual orientation, creed, marital status, political affiliation, personal appearance, or on the basis of rights secured by the First Amendment, in all aspects of employment, educational programs and activities, and admissions.

To apply, applicants are required to submit a resume and cover letter to Dr. Robert D. Lamb, Director of ICONS Project, at rdlamb@umd.edu. 

Contact: Robert D. Lamb, Director of ICONS, rdlamb@umd.edu
Required Documents: Resume, cover letter
Salary: $25 to $35 per hour, depending on experience