Machine Learning Research Scientist - Frontier Lab

Software Engineering Institute | Carnegie Mellon University • Pittsburgh, PA

23h • Onsite

About The Position

At the SEI AI Division, we conduct research in applied artificial intelligence and the engineering challenges related to building, deploying, and sustaining AI-enabled systems for high-impact government missions. The Frontier Lab advances AI engineering and transitions frontier AI capabilities to government stakeholders through applied research, rapid prototyping, short-cycle test and evaluation, and technical advisory. Position Summary As a Machine Learning Research Scientist in the Frontier Lab, you will conduct applied AI/ML research and develop prototype capabilities that inform and improve real government and DoW workflows. You will execute work in mission context—developing an appreciation for users, operational constraints, and intended outcomes—and translate sponsor needs into technically credible approaches and evidence. This role spans the research-engineering spectrum: some MLRS hires may lean more research-heavy and others more engineering-heavy, but successful candidates collaborate effectively across both. Frontier Lab work spans several complementary focus areas, including: Agentic AI for mission workflows (e.g., planning, analysis, decision support) where autonomous and human-guided agents interact with tools, data systems, and operators. AI test, evaluation, verification, and validation (TEVV) to improve confidence in performance, robustness, uncertainty, and trustworthiness of ML-enabled systems. Mission-tailored language models, including techniques to improve accuracy and reliability, reduce hallucinations, and integrate structured knowledge for operational tasks. Mission modalities and multimodal learning, including sensor fusion and learning under noisy, sparse, or constrained data conditions (including synthetic data and weakly-/self-supervised approaches). AI at the tactical edge, enabling capability under constrained compute/connectivity through efficient inference, compression, rapid adaptation, and update/redeploy patterns.

Requirements

• BS in Electrical Engineering, Computer Science, Statistics, or related discipline with eight (8) years of experience in hands-on software development; OR MS in the same fields with five (5) years of experience; OR PhD with two (2) years of relevant experience.
• Strong foundation in machine learning and statistical learning, including experiment design and evaluation.
• Demonstrated ability to implement ML systems in Python using modern ML libraries (e.g., PyTorch / TensorFlow) and common scientific tooling.
• Demonstrated ability to communicate technical results clearly in written deliverables and presentations.
• Ability to work effectively with ambiguity and deliver results in iterative project cycles with strong self-direction.
• Explains technical content clearly; translates between mission problems and technical approaches.
• Designs sound experiments; recognizes evaluation pitfalls (leakage, confounds, distribution shift).
• Balances research quality with timelines and constraints; produces credible evidence and useful prototypes.
• Works well in interdisciplinary teams; contributes effectively to shared code and shared evaluation approaches.
• Executes independently with low oversight; manages time effectively; escalates risks early and seeks guidance when needed.
• Flexible to travel to SEI offices in Pittsburgh, PA and Washington, DC / Arlington, VA, sponsor sites, conferences, and offsite meetings (~10% travel).
• You must be able and willing to work onsite at an SEI office in Pittsburgh, PA or Arlington, VA 5 days per week.
• You will be subject to a background investigation and must be able to obtain and maintain a Department of War security clearance.

Nice To Haves

• Applied ML research and prototyping for real operational workflows, including tool-integrated AI systems and human-in-the-loop settings.
• Designing and operating evaluation pipelines for LLMs and/or CV models (benchmarking, regression testing, robustness checks, scenario-based evaluations).
• Language model grounding and reliability techniques (structured knowledge integration, RAG, tool use, error analysis).
• Learning under constrained/noisy data conditions (synthetic data, programmatic labeling, semi-/self-supervised learning).
• Edge-relevant ML (compression, quantization, distillation, efficient inference, rapid adaptation patterns).
• Evidence of research output: publications, technical reports, open-source contributions, or applied research artifacts.
• Experience working with government/DoW stakeholders or in high-assurance environments.

Responsibilities

• Mission-context execution: Execute tasks within the mission context, considering users, use cases, operational constraints, and intended outcomes. Translate sponsor goals into clear technical questions, measurable success criteria, and credible evaluation evidence.
• Applied research and experimentation: Design and conduct studies grounded in mission needs; form hypotheses, run controlled experiments, analyze results, and produce actionable recommendations.
• Prototype capability development: Build research prototypes, evaluation harnesses, and reference implementations that demonstrate feasibility and generate learning in realistic settings.
• Evaluation and assurance (TEVV): Develop and apply evaluation methodologies for ML systems (especially CV and LLMs), including metrics, benchmark design, robustness testing, uncertainty and calibration approaches, and repeatable test pipelines.
• Engineering rigor appropriate to the task: Write clear, maintainable code and documentation with a level of engineering discipline proportionate to the intended use. Emphasize reproducibility and handoff-ready artifacts suitable for downstream integration and operational hardening through formal DevSecOps processes.
• Iterative execution, self-direction, and time management: Plan and deliver work in iterative cycles; manage priorities effectively; communicate status and risks early; and maintain momentum with minimal supervision.
• Customer translation and communication: Communicate technical progress and results clearly to technical and non-technical stakeholders through briefings, demos, reports, and recommendations.
• Publication and knowledge dissemination: Identify opportunities to publish research insights and lessons learned at reputable venues (e.g., NeurIPS, ICLR, MLCON, etc.), subject to customer and releasability constraints.
• Team collaboration: Contribute to technical discussions shaping tasking and delegation, support shared project goals, and provide guidance to junior teammates when appropriate.

What This Job Offers

• Job Type: Full-time
• Career Level: Mid Level
• Education Level: Ph.D. or professional degree
• Number of Employees: 5,001-10,000 employees