Perception Engineer - Multimodal and Embodied AI

Shield; San Diego (CA); Vor Ort; USD 120.000 - 160.000; Vollzeit

Zusammenfassung

Shield AI is seeking a perception engineer to translate AI research into real autonomous systems. You will evaluate multimodal perception methods, build prototypes, and contribute to Hivemind perception capabilities deployed on edge robotics hardware.

This role emphasizes robust experiments, sensor constraints, edge deployment, documentation, and collaboration with autonomy teams and product engineers to deliver production-ready perception solutions.

Qualifikationen

• Research depth in computer vision, robotics, machine learning, or embodied learning.
• Experience with multimodal vision, visual foundation models, video understanding, or open-vocabulary perception.
• Strong Python skills and practical work with PyTorch or a similar machine learning framework.
• Ability to design careful experiments, understand model failures, and build substantial research software.
• Interest in production-quality C++ and collaboration with product engineering teams.

Aufgaben

• Evaluate emerging perception and embodied-AI methods against representative autonomy data.
• Build prototypes that expose where models work, fail, and can be deployed.
• Combine learned and classical approaches for detection, tracking, geometric vision, or sensor fusion.
• Optimize and integrate perception capabilities for edge robotics systems.
• Carry successful experiments into tested Hivemind capabilities that other teams can adopt.

Kenntnisse

Python

Experiment design

Research in CV

Research software

Tools

PyTorch

C++

Jobbeschreibung

What the role actually is

Shield AI is hiring a perception engineer to move promising AI research into real autonomous systems. The role evaluates multimodal computer vision, visual foundation models, and embodied AI methods, then turns the useful pieces into Hivemind perception capabilities.

This is a research-to-product role: experiments and prototypes matter, but so do sensor constraints, edge deployment, testing, documentation, and adoption by autonomy teams.

What you would work on

• Evaluate emerging perception and embodied-AI methods against representative autonomy data
• Build prototypes that expose where models work, fail, and can be deployed
• Combine learned and classical approaches for detection, tracking, geometric vision, or sensor fusion
• Optimize and integrate perception capabilities for edge robotics systems
• Carry successful experiments into tested Hivemind capabilities that other teams can adopt

What they are asking for

• Research depth in computer vision, robotics, machine learning, or embodied learning
• Experience with multimodal vision, visual foundation models, video understanding, or open-vocabulary perception
• Strong Python skills and practical work with PyTorch or a similar machine learning framework
• Ability to design careful experiments, understand model failures, and build substantial research software
• Interest in production-quality C++ and collaboration with product engineering teams

Why this one is worth a look

This is one of the cleaner bridges between the current foundation-model wave and deployed robotics. The caveat is that it sits inside defense autonomy, but the technical problem is exactly the one many robot companies are wrestling with: making general AI perception useful on real hardware under real constraints.