Principal / Sr. Principal AI Software Engineer

Northrop Grumman Northrop Grumman · Aerospace · Dulles, VA +2 · Software

AI Software Engineer focused on Reinforcement Learning (RL) and Supervised Learning (SL) for space and aerospace programs. Responsibilities include novel algorithm R&D, designing and implementing RL/SL algorithms, prototyping in Python/JAX/PyTorch, porting to C++/CUDA, developing physics-based autonomy for mission planning and decision-making, building models for constrained timeline re-optimization, and developing AI solutions for real-time anomaly detection and response. The role also involves leading verification and flight readiness campaigns (Monte-Carlo, HIL, DIL) and participating in the full software development lifecycle. The position requires a Top Secret clearance and experience with AI/ML model development, particularly RL or SL.

What you'd actually do

  1. Perform Novel Algorithm R&D
  2. Design and implement state-of-the-art RL / SL algorithms drawn from the latest literature
  3. Rapidly prototype in Python/JAX/PyTorch, then port to embedded C++/CUDA
  4. Develop Physics-Based Autonomy to perform Mission Planning & Decision-Making
  5. Apply supervised learning, reinforcement learning, and other AI/ML techniques to high-fidelity astrodynamics planning and controls problems, including real-time constraint handling

Skills

Required

  • Bachelor's degree with a minimum of 5 years of relevant AI engineering experience (or equivalent experience)
  • Master's degree with a minimum of 3 years of relevant AI engineering experience (or equivalent experience)
  • PhD with a minimum of 1 year of relevant AI engineering experience (or equivalent experience)
  • Industry knowledge and/or foundational education of AI, with a focus on ML, RL, or SL model development
  • Experience with machine learning
  • Python
  • JAX
  • PyTorch
  • C++
  • CUDA
  • Top Secret security clearance
  • US citizenship

Nice to have

  • Deep knowledge of Computer Science
  • Deep focus on Reinforcement Learning (RL)
  • Physics-Based Autonomy
  • Mission Planning & Decision-Making
  • astrodynamics planning and controls problems
  • real-time constraint handling
  • Fuse learned policies with classical GNC filters
  • Build models that re-optimize delta-V, power, and comm- constrained timelines
  • neural search
  • differentiable optimization
  • AI solutions for real-time anomaly detection and response
  • detect out-of-family telemetry
  • hierarchical or policy-gradient RL
  • Lead Verification & Flight Readiness
  • Monte-Carlo
  • Processor-in-the-Loop
  • Hardware-in-the-Loop
  • digital twin campaigns
  • full life cycle of software development
  • requirements development
  • modeling and design
  • application development
  • unit to CSCI testing
  • integration
  • formal system testing
  • release, installation, and maintenance
  • Sensitive Compartmented Information (SCI)/Special Access Program (SAP) approval/access

What the JD emphasized

  • Top-Secret (TS) security clearance
  • Reinforcement Learning (RL)
  • supervised learning
  • AI/ML techniques

Other signals

  • Reinforcement Learning
  • Autonomy
  • Spacecraft Operations
  • Real-time Anomaly Detection
  • Mission Planning