وظيفة Pipefitter Journeyman واحدة قريبًا مني

About Durandal

At Durandal, quality isn’t just a standard, it is part of who we are. We take pride in developing industry-leading technology across four core areas of expertise: Software Engineering, Modeling and Simulation, Electronic Warfare, and AI/ML Engineering. Our talented team of developers and engineers brings curiosity, expertise, and a shared commitment to excellence in everything we do. Together, we challenge ourselves to build software and technology that is powerful, accurate, intuitive, and dependable, solving complex problems and making a real difference for our customers. We celebrate the teamwork, innovation, and attention to detail that make it all possible.

Join our team and become part of a talented, collaborative, and driven group of people who are passionate about what we do. At Durandal, you’ll have the opportunity to make an impact, grow your skills, tackle meaningful challenges, and help shape our future technology.

Position Summary

Durandal is hiring a journeyman AI/ML engineer to build, evaluate, and field the learning systems at its core. Cognitive EW puts machine learning directly in the RF loop: characterizing unknown emitters, reasoning about the spectrum in real time, and selecting or synthesizing responses under hard latency, power, and size constraints. That problem set looks very little like natural language work, and we are staffing accordingly.

The engineer in this role will work across the full lifecycle - generating synthetic RF training data, training models against large collected and simulated datasets, adjudicating competing model candidates, validating that simulation performance carries over to live hardware, and deploying to embedded targets. Expect to sit between the signal processing engineers, the M&S team, and the test community rather than inside a single lane.

Core Responsibilities

Modeling & Algorithm Development

• Design, train, and tune models for emitter characterization, signal classification, anomaly and novelty detection, and technique selection against RF and IQ data.
• Work primarily in model families other than LLMs and transformers; convolutional and residual architectures, recurrent and temporal convolution networks, state-space models, graph neural networks, Bayesian and probabilistic methods, Gaussian processes, reinforcement and online learning, autoencoders, and classical statistical learning, and justify the architecture choice on the merits of the signal problem.
• Implement custom loss functions, augmentation pipelines, and training objectives suited to RF data characteristics such as low SNR, class imbalance, non-stationarity, and severely limited labeled truth.

Signal Processing & EW Integration

• Build feature extraction and preprocessing chains over IQ streams, spectrograms, and time-frequency representations, and reason about the tradeoff between learned features and conventional DSP front ends.
• Collaborate with EW subject matter experts on electronic support, electronic attack, and electronic protection use cases, threat emitter behavior models, and technique effectiveness criteria.
• Integrate models with SDR and RF hardware chains, and account for receiver artifacts, calibration drift, and channel effects in model design.

Synthetic Data & Simulation-to-Live Congruency

• Develop synthetic RF data generation pipelines, including parametric scenario generation, channel and propagation modeling, domain randomization, and generative augmentation.
• Quantify and close the gap between simulated and live performance: measure domain shift, build congruency metrics, and drive changes to either the simulation fidelity or the model as the evidence indicates.
• Support hardware-in-the-loop and range test events, including pre-test prediction, live data capture, and post-test reconciliation against simulated expectations.

Model Adjudication, Test & Evaluation

• Stand up model adjudication frameworks that compare candidate models across scenarios and select, arbitrate, or ensemble among them at runtime.
• Define evaluation methodology and acceptance criteria: held-out and out-of-distribution test sets, ablation studies, confidence calibration, failure mode characterization, and regression testing across model versions.
• Produce the technical evidence and documentation that testers, program leadership, and the customer need to trust a model in an operational decision loop.

Training Infrastructure & Deployment

• Manage large-scale dataset curation, labeling workflows, versioning, and storage for multi-terabyte RF corpora.
• Execute distributed and multi-GPU training regimes, including hyperparameter search, curriculum and staged training, checkpointing, and experiment tracking.
• Optimize and deploy models to production and embedded targets under SWaP constraints — quantization, pruning, distillation, graph export, and runtime integration into C++ processing chains.

Required Qualifications

• MUST BE U.S. CITIZEN
• Education: B.S. in Computer Science, Computer Engineering, or Electrical Engineering.
• Experience: 3–10 years of professional experience in machine learning, signal processing, or a closely related engineering discipline.
• Python: Strong Python proficiency with at least one major ML framework (PyTorch, JAX, Keras/TensorFlow, or equivalent) and the surrounding scientific stack.
• C++: Demonstrated production C++ experience, especially related to high speed computer and compiled software deployment.
• Model breadth: Demonstrated depth in model families beyond LLMs and transformers, with the ability to explain why a given architecture fits a given signal problem.
• Training at scale: Hands-on experience training on large datasets, including data pipeline construction, GPU training, and systematic experiment management.
• Evaluation rigor: Experience designing test and evaluation approaches for models, not only reporting aggregate accuracy on a benchmark split.
• Collaboration: Ability to work directly with domain engineers and testers and to communicate model behavior and limitations to a non-ML audience.

Preferred Qualifications:

• M.S. in Computer Science, Computer Engineering, Electrical Engineering, or a related technical field.
• Direct experience in electronic warfare, SIGINT, ELINT, radar, or spectrum operations.
• Digital signal processing background: filtering, detection and estimation, direction finding, beamforming, modulation recognition, or emitter fingerprinting.
• Synthetic data generation for sensor or RF domains, including generative models used for augmentation rather than content.
• Reinforcement learning or online/adaptive learning applied to closed-loop control or technique selection.
• MLOps tooling for containerized training and deployment, CI/CD for models, and reproducible experiment infrastructure.
• Prior work on DoD programs and familiarity with the associated documentation, review, and accreditation processes.
• Publications, open-source contributions, or competition results in RF machine learning.

Pay: $100,000.00 - $180,000.00 per year

Benefits:

• 401(k)
• Dental insurance
• Health insurance
• Life insurance
• Relocation assistance
• Retirement plan
• Vision insurance

Work Location: In person