Applied Researcher: On-Device Multimodal Reasoning

Socket.dev; Sunnyvale (CA); On-site; USD 200,000 - 320,000; Full time

Job summary

Apple is seeking an Applied Researcher to advance on-device multimodal reasoning for vision-language systems. The role focuses on compact VLM design that can reason over images, video, and 3D content within strict compute and memory budgets on Apple hardware.

You will partner with hardware, compiler, and ML teams to push the boundaries of small-model reasoning, enabling real-time multimodal understanding while preserving privacy and offline capability.

Qualifications

• MS in CS/ML/CV or related field or equivalent practical experience.
• Strong foundation in deep learning with experience in LLM or VLM training, post-training, or inference optimization.
• Demonstrated experience with multimodal models in resource-constrained environments, including reasoning quality, decoding, or model compression.

Responsibilities

• Own the reasoning component of the on-device multimodal stack, designing compact VLMs for on-device use.
• Collaborate with hardware, compiler, and ML researchers to optimize inference under tight compute and memory budgets.
• Research and implement efficient decoding, quantization, distillation, and KV-cache compression for real-time performance.

Skills

Python

PyTorch

Multimodal models

Model compression

Education

MS in CS/ML/CV or related field

PhD (preferred)

Tools

CoreML

MLX

TensorRT-LLM

llama.cpp

vLLM

Job description

Apple is seeking an Applied Researcher to advance on-device multimodal reasoning for vision-language systems. The role focuses on compact VLM design that can reason over images, video, and 3D content within strict compute and memory budgets on Apple hardware.

You will partner with hardware, compiler, and ML teams to push the boundaries of small-model reasoning, enabling real-time multimodal understanding while preserving privacy and offline capability.