Research Engineer - Perception and Machine Learning

Meta Meta · Big Tech · Bellevue, WA +1

Research Engineer focused on bringing AI research (CV, Embodied AI, Multimodal LLMs) to power-constrained edge devices like smart glasses and robots. Responsibilities include system architecture, end-to-end ownership, experimental design, cross-functional collaboration, and potentially technical leadership for staff-level hires. Requires strong C++/Python, PyTorch/TensorFlow, ML deployment experience, and a publication record in relevant fields.

What you'd actually do

  1. Translate cutting-edge research in Vision-Language Models (VLMs), Reinforcement Learning, and Multimodal LLMs into performant, real-world applications on smart glasses and robotic platforms
  2. Own the full lifecycle of feature development, from initial prototyping and data collection to deployment and system integration
  3. Design and lead large-scale ablation studies; develop robust benchmarking suites to evaluate and iterate on next-gen contextual AI
  4. Partner with Hardware Engineers to influence sensor/silicon design and collaborate with Researchers and Product Managers to define the future of human-AI interaction
  5. When hired at a staff level, lead the design and execution of engineering roadmaps, making critical architectural decisions to ensure low-latency, high-accuracy inference on power-constrained "always-on" edge devices.

Skills

Required

  • C++
  • Python
  • PyTorch or TensorFlow
  • model optimization (e.g., quantization, distillation, or custom kernel development)
  • deploying machine learning models into production environments or integrated hardware-software systems
  • high-dimensionality, multi-modal datasets
  • publications at top-tier venues (CVPR, ICCV, NeurIPS, ICRA, RSS) or significant patent filings
  • Egocentric Perception
  • Vision-Language-Action (VLA) models
  • SLAM
  • Sim-to-Real transfer

Nice to have

  • Ph.D. or M.S. in Computer Science, Software Engineering, or Robotics
  • Reinforcement Learning
  • Multimodal LLMs
  • technical leadership and mentorship

What the JD emphasized

  • power-constrained, egocentric devices
  • low-latency, high-accuracy inference on power-constrained "always-on" edge devices

Other signals

  • Move beyond algorithm implementation to architecting the systems that power always-on contextual AI.
  • Drive the engineering efforts to bring research breakthroughs from high-compute clusters to power-constrained, egocentric devices like smart glasses and robotic platforms.
  • Own the full lifecycle of feature development, from initial prototyping and data collection to deployment and system integration.
  • Design and lead large-scale ablation studies; develop robust benchmarking suites to evaluate and iterate on next-gen contextual AI.
  • When hired at a staff level, lead the design and execution of engineering roadmaps, making critical architectural decisions to ensure low-latency, high-accuracy inference on power-constrained "always-on" edge devices.