Senior ML Engineer - On-Device for Mobile & Embedded

Socket.dev; Mountain View (CA); On-site; USD 207,000 - 300,000; Full time

Job summary

Google engineers are sought to design and deliver the real-time on-device machine learning foundation, turning raw camera and sensor input into actionable signals for mobile apps. You will optimize models for NPU, GPU, and DSP execution, including quantization and hardware-aware design.

Own models end-to-end from prototype to deployment on hundreds of millions of devices, with training in JAX/TensorFlow and C++ integration into the camera pipeline.

Qualifications

• Bachelor's degree or equivalent practical experience required.
• 8 years of software design and architecture experience.
• Experience with C, C++, machine learning, and embedded systems.
• Experience with machine learning algorithms.
• Experience with machine learning architecture and ML research.

Responsibilities

• Design and implement the real-time on-device ML foundation for camera and sensor input.
• Optimize ML models for NPU, GPU, and DSP with quantization and hardware-aware design.
• Own models end-to-end from prototype to deployment on hundreds of millions of devices.
• Collaborate with product, UX, software, and hardware teams to define requirements.
• Establish best practices for ML development, deployment, and evaluation.

Skills

8 years software design & architecture

C

C++

Machine learning

Embedded systems

ML algorithms

ML architecture

ML research

ML production on mobile/embedded

Quantization / hardware-aware design

Custom kernels

JAX/TensorFlow

Education

Bachelor's degree or equivalent practical experience

Master’s degree or PhD in Engineering, CS, or related field

Tools

JAX/TensorFlow

Job description

Google engineers are sought to design and deliver the real-time on-device machine learning foundation, turning raw camera and sensor input into actionable signals for mobile apps. You will optimize models for NPU, GPU, and DSP execution, including quantization and hardware-aware design.

Own models end-to-end from prototype to deployment on hundreds of millions of devices, with training in JAX/TensorFlow and C++ integration into the camera pipeline.