Senior Staff Software Engineer, Perception (R4985)

Shield AI•Washington, DC

4d•$233,760 - $350,640

About The Position

The Hivemind Solutions Perception team develops the next generation of perception capabilities for autonomous systems by combining state-of-the-art machine learning with the proven foundations of computer vision. The team advances how autonomous platforms understand and interpret the world by developing vision, vision-language (VLM), and vision-language-action (VLA) models that tackle core perception challenges such as object understanding, scene interpretation, and mission-relevant environmental awareness. Working at the intersection of research and production, our engineers build the data pipelines, supervised fine-tuning (SFT) workflows, evaluation frameworks, and deployment infrastructure needed to transform cutting-edge AI research into reliable, mission-ready perception capabilities. In this role, you'll develop and deploy advanced machine learning models that enable autonomous systems to better understand, reason about, and interact with their environment. You'll partner closely with machine learning researchers, autonomy engineers, perception engineers, and platform teams to translate emerging AI capabilities into reliable, production-ready systems for U.S. and international defense customers. This is an ideal opportunity for engineers who enjoy building state-of-the-art AI systems while solving the practical challenges of deploying them on operational autonomous platforms.

Requirements

Typically requires a minimum of 10 years of related experience with a Bachelor’s degree; or 9 years and a Master’s degree; or 7 years with a PhD; or equivalent work experience.
Mastery of machine learning fundamentals.
Experience training an deploying ML models for computer vision in a production setting.
Strong understanding of 3D vision problems/algorithms.
Experience with machine learning frameworks such as PyTorch and TensorFlow.
Demonstrated expertise in deploying models using TensorRT and ONNX.
Proficiency in C++ and Python.
Strong analytical and problem-solving skills, with the ability to translate research into practical applications.
Ability to obtain a SECRET clearance

Nice To Haves

Experience with developing autonomous systems for defense customers.
Experience with training/finetuning vision-language models, vision-language-action models, and/or world models.
Contributions to open-source projects in machine learning or computer vision.
Track record of publications in leading computer vision and robotics conferences and journals (e.g., CVPR, ICCV/ECCV, RAL, ICRA).

Responsibilities

Model Development – Design, train, fine-tune, and maintain state-of-the-art vision, vision-language, and vision-language-action models that improve perception and decision-making for autonomous systems.
Data Pipelines & Model Training – Build scalable data pipelines, supervised fine-tuning (SFT) workflows, and evaluation loops that continuously improve model performance on mission-relevant tasks.
Model Deployment & Optimization – Deploy and optimize machine learning models for embedded hardware using technologies such as ONNX, TensorRT, and hardware-accelerated inference frameworks.
Perception & Autonomy Applications – Apply modern machine learning techniques to solve challenging perception and autonomy problems across aerial and other autonomous systems operating in complex, real-world environments.
Research-to-Production – Translate cutting-edge machine learning research into production-ready capabilities by balancing model performance, robustness, computational efficiency, and operational reliability.
Cross-functional Collaboration – Partner closely with perception, autonomy, platform, and software engineering teams to integrate machine learning capabilities into mission-ready autonomous systems.
Model Evaluation & Validation – Develop benchmarks, testing methodologies, and evaluation frameworks to measure model performance, identify failure modes, and guide future improvements.
Continuous Improvement – Improve training infrastructure, developer tooling, deployment workflows, and model lifecycle management to accelerate experimentation and production delivery.