Intermediate AI and Decision Engineer

What matters most

• Build decision-support prototypes around real workflows

  The role develops and fields AI-enhanced software prototypes based on workflow and operational needs. Prepare to explain how you would translate a decision-maker’s workflow into a prototype and test whether it supports the intended decision.

• Evaluate against sponsor requirements

  MITRE calls for evaluating AI solutions against sponsor requirements and producing actionable recommendations for consequential decisions. Prepare to describe how you would define evaluation criteria, examine limitations, and communicate a recommendation.

• Apply machine learning and LLMs to decision problems

  The posting requires hands-on machine learning, including deep learning and reinforcement learning, and describes applying LLMs to sponsors’ challenges. Prepare to explain when a particular approach would help a decision task—and when it would not.

• Advance research with interdisciplinary teams

  Responsibilities include proposing and executing research, identifying R&D opportunities, and leading projects or small teams. Prepare an example of turning a technical gap into a focused research plan while explaining the work to both expert and nonexpert collaborators.

Likely challenges to think through

• Balancing usefulness with high-stakes limitations

  A likely difficulty is making AI output useful without overstating its reliability in a consequential decision. Prepare to discuss failure cases, evaluation evidence, and how a system could make uncertainty visible to its users; this is a preparation hypothesis, not a reported MITRE problem.

• Adapting methods across sponsor workflows

  Because the role spans interdisciplinary teams and sponsors’ decision problems, a plausible trade-off is between reusing an AI technique and tailoring it to a specific operational workflow. Prepare to describe how you would learn the workflow before selecting or changing an approach.

• Connecting novel research to a fieldable prototype

  The posting combines cutting-edge research with developing and fielding prototypes. A likely challenge is choosing how much novelty to pursue while still producing something evaluable against sponsor requirements. Prepare to explain how you would scope experiments and decide what evidence warrants further development.

Topics to prepare

• Machine learning, deep learning, and reinforcement learning

  Review when each approach fits a decision problem, what feedback it needs, and how failure modes differ. Suggested exercise: compare a supervised model, a deep model, and a reinforcement-learning policy for one hypothetical decision scenario. Hands-on experience across these areas is a basic qualification.

• High-stakes decision support

  Review how operational workflows, uncertainty, and human judgment affect AI-assisted decisions. Suggested exercise: map a hypothetical workflow, identify where an AI recommendation could fail, and state what a decision-maker would need to know. The posting explicitly calls for identifying gaps in high-stakes AI decision-making.

• Large language models in decision systems

  Review suitable uses and limitations of large language models for problem-solving and decision support. Suggested exercise: outline an LLM-assisted workflow and mark where its output needs checking before action. Applying LLMs is part of the role; recent LLM experience is preferred.

• Evaluating AI against sponsor requirements

  Review how requirements become evaluation criteria and how results support an actionable recommendation. Suggested exercise: write three criteria for a hypothetical decision-support system and explain what evidence would support a recommendation. Evaluation against sponsor requirements is a stated responsibility; explainable AI is preferred, not required.

Skills to revise

• Programming a small AI component

  Suggested exercise: implement a simple model or decision component in Python or C++, then explain its inputs, outputs, and limitations. Programming experience with languages such as Python and C++ is required.

• Building a workflow-informed prototype

  Suggested exercise: sketch and implement a minimal decision-support prototype for a hypothetical workflow, documenting which operational need each feature addresses. The role develops prototypes based on workflow and operational analysis; prototype development and deployment experience is preferred.

• Running a focused evaluation

  Suggested exercise: test a hypothetical AI prototype against stated requirements, record failure cases, and write a recommendation grounded in the results. The posting calls for evaluating AI solutions and producing actionable recommendations.

• Explaining trade-offs to different audiences

  Suggested exercise: present the same technical recommendation in a brief for an expert and a plain-language brief for a nonexpert, including its uncertainty and limitations. Written and verbal communication with both audiences is a basic qualification.

• Scoping applied research with a team

  Suggested exercise: draft a one-page proposal identifying a decision-system gap, a testable research question, and how a small cross-functional team would investigate it. The role includes proposing and executing research and leading projects or small teams.

• Build a requirements evidence sheet

  Map the posting’s degree and experience paths, hands-on machine learning (including deep learning and reinforcement learning), one listed specialty, and Python or C++ experience to examples you can substantiate. Mark any gaps rather than inventing evidence.

• Prepare a decision-system project walkthrough

  If you have a relevant project, outline the workflow need, prototype design, your contribution, evaluation, and what was fielded or learned. The role calls for decision-support prototypes shaped by operational needs; use a hypothetical design instead if you lack a comparable project.

• Practice a high-stakes design exercise

  As interview preparation, sketch how you would turn a consequential decision workflow into an AI-supported prototype. State the human decision points, assumptions, failure modes, and where an LLM would or would not help.

• Draft an evaluation and recommendation brief

  Choose a hypothetical sponsor requirement and describe test cases, measures, limitations, and a recommendation based on results. This mirrors the posting’s responsibility to evaluate AI solutions against sponsor requirements.

• Outline a small research proposal

  Prepare a one-page idea for an AI-enhanced decision-system research question, a feasible first experiment, and criteria for continuing. Be ready to explain how you would coordinate work across disciplines or a small team.

• Rehearse two versions of your explanation

  Explain the same technical trade-off briefly to an AI specialist and to a nontechnical decision-maker. The posting explicitly calls for communication with both audiences and for identifying gaps in high-stakes AI decision-making.

• Prepare practical and technical questions

  Ask how sponsors define decision requirements, how prototypes are evaluated and fielded, and what an intermediate engineer would lead. Confirm the applicable onsite location and clearance timing: the posting requires at least three onsite days weekly, U.S. citizenship for clearance eligibility, and a Secret clearance within one year of hire, but not at application.