AI / LLM & ML Engineer

UIC Alaska • San Diego, CA

68d

About The Position

CYBER SECURITY ENGINEER (HITS-R): Bowhead seeks a part-time (10-20 hours a week) AI / LLM & ML Engineer to support the High Performance Computing Modernization Program (HPCMP) Integrated Technical Services -Restricted (HITS-R) contract. Bowhead will provide the High Performance Computing Modernization Program (HPCMP) with technical and professional support elements required for the operation of the HPCMP Office (HPCMPO) in all phases of planning and execution of its mission, including: technical and administrative support in all areas of program activity, program management support, meeting facilities and office environment, to include a complete telecommunications and computer systems capability and full logistical services. The AI / LLM & ML Engineer to help build fenced-off , HPC-aware AI systems that sit directly in the CREATE ecosystem (Kestrel, Helios, Sage, etc.). You will design and deploy agentic LLM workflows, RAG systems, and surrogate models that accelerate physics-based simulations, automate input setup, and improve developer productivity on GPU-enabled Lambda workstations and DoD-style HPC systems. This role is ideal for advanced undergraduate student in AI/ML/CS/CE who already has significant prior hands-on experience with LLMs, RAG, Linux, and GPU/HPC environments, and is excited to work at the intersection of CFD/physics codes, surrogate modeling, and AI-enabled workflows.

Requirements

• Education: Currently pursuing a degree in Computer Science, Computer Engineering, Data Science, Applied Math, or a closely related field with focus in AI/ML.
• Core Technical Skills: Strong programming experience in Python (PyTorch and/or TensorFlow; Hugging Face or similar LLM ecosystems).
• Prior, hands-on experience building or fine-tuning LLMs and implementing RAG pipelines (vector stores, embeddings, retrieval, prompt orchestration).
• Prior experience in building agentic AI models and Large Action models
• Extensive hand-on experience with Linux environments (shell, ssh, basic admin, environment management).
• Experience working with GPUs (e.g., CUDA-based training/inference, profiling basic performance).
• HPC & CREATE-Related Experience: Demonstrated experience running workloads on HPC systems (job schedulers, batch scripts, scaling basics).
• Prior exposure to CREATE codes (Helios and/or Kestrel), or equivalent large-scale CFD / physics-based solvers, including basic input deck structure and run workflows.
• Familiarity with AI surrogate modeling concepts (e.g., DNNs, CNNs, surrogates, reduced-order models).
• Soft Skills: Ability to work independently, ask good technical questions, and iteratively refine solutions with domain experts.
• Strong written communication skills for documenting workflows, APIs, and architectural decisions.
• Must be able to lift up to 25 pounds
• Must be able to stand and walk for prolonged amounts of time
• Must be able to twist, bend, and squat periodically
• Must currently hold a security clearance at the Top Secret level, may be required to obtain a Top Secret/SCI clearance upon hire.
• US Citizenship is a requirement for Top Secret clearance at this location.

Responsibilities

• Fenced-Off Agentic LLM & RAG Systems Design and implement “fenced-off” (isolated, secure) LLM-based assistants to support: Auto physics input setup and configuration checking for Helios/Kestrel and related CREATE workflows. End-user guidance for simulation setup, run-time control, and post-processing. Build RAG pipelines over CREATE documentation, examples, templates, and historical cases to enable trustworthy, domain-aware responses. Implement guardrails, schema validation, and automated tests so LLM agents generate physically consistent and executable input decks (e.g., for CFD, rotorcraft, EM, or mission models).
• Auto Physics Input Setup & Evaluation Develop tools that automatically: Ingest user requirements and baseline geometries/conditions. Generate candidate input configurations for CREATE solvers (Helios/Kestrel/etc.). Run sanity checks against best practices and project constraints before submission to HPC queues. Integrate these tools with existing workflow automation / job submission systems on Lambda GPU boxes and, where appropriate, remote HPC centers.
• Internal AI “Vibe Coding” / Developer-Assist Infrastructure Prototype and evaluate internal “AI coding copilot” workflows for CREATE teams: Agentic LLMs for refactoring, unit test generation, and boilerplate code generation. AI-enabled DevOps concepts (CI support, log triage, failure summarization) where feasible. Help define best practices for AI-assisted development in a secure, fenced-off environment (no external cloud dependencies on production data/code).
• GPU-Enabled Speedups for Physics-Based Codes Collaborate with domain experts to implement ML/LLM components that: Accelerate solver workflows (e.g., better initial guesses, reduced-order models, or auto-generated surrogate models). Support fast design-space exploration and parametric studies using Helios/Kestrel and Sage. Benchmark and profile GPU-accelerated solutions on Lambda systems to quantify speedups vs. CPU or baseline runs.
• Surrogate Modeling & Data Pipelines Work with existing CREATE surrogate modeling tools (e.g., Sage) to: Build data pipelines from large CFD/physics simulations into training datasets. Train, validate, and package surrogate models for real-time or near real-time predictions. Assist in integrating physics-informed or physics-incorporated ML models into broader workflows for vehicle design and mission analysis.

What This Job Offers

• Job Type: Part-time
• Career Level: Intern