below marketmost similar roles pay hereabove market
This role pays less than
52%
of similar roles. Most pay
$196,633–$254,750
— the blue band above.
At the midpoint, this role pays about
$214k
versus about
$226k
for comparable roles.
Based on 240 similar postings.
Employer
About Apple Inc
Apple Inc. is a multinational technology company known for designing and manufacturing consumer electronics, software, and online services, including the iPhone, Mac, iPad, and App Store. Industry: Consumer Electronics & Software
Apple Inc currently has
3497 open roles
on FindRole.
Listed pay typically runs
$174,149–$277,600
across 2702 roles with salary data.
The Applied Researcher: On-Device Multimodal Reasoning joins the Video Computer Vision organization to develop next-generation multimodal AI systems for Apple products. You will design, train, and post-train compact vision-language models under 10B parameters to perform multi-step visual reasoning within strict compute, memory, and latency constraints. Key responsibilities include implementing compressed chain-of-thought, reasoning distillation, and hardware-aware inference optimizations like speculative decoding and quantization. You will also build systems that reason over real-time perception experts, such as 3D scene reconstruction and human pose recovery. The role requires proficiency in Python and PyTorch, with expertise in VLM architectures, RL methods like GRPO, and optimization toolchains including MLX and CoreML. This work solves the technical challenge of enabling high-quality, privacy-preserving, on-device multimodal understanding without relying on cloud connectivity.
Design and train compact vision-language models under 10B parameters for multi-step visual reasoning on-device.
Implement efficient reasoning techniques like compressed chain-of-thought, reasoning-trace distillation, and adaptive test-time compute.
Own the on-device decoding stack including speculative decoding, KV-cache compression, and quantization.
Integrate real-time perception experts like 3D scene reconstruction and human pose recovery into the VLM reasoning path.
Develop structured-output fusion and scheduling policies for perception models to manage resolution and frame rate budgets.
Apply reasoning-focused post-training using supervised distillation, RL methods like GRPO, and process supervision.
Optimize model architectures for Apple silicon to minimize latency, memory footprint, and power consumption.
Build evaluation frameworks to measure reasoning fidelity and spatial grounding against fixed device budgets.
What we're looking for
MS in Computer Science, Machine Learning, AI, Computer Vision, or a related field (or equivalent practical experience).
Strong foundation in deep learning with specific experience in LLM or VLM training, post-training, or inference optimization.
Demonstrated experience working with multimodal models in resource-constrained environments, including reasoning quality, decoding, or model compression.
Proficiency in Python and modern deep learning frameworks (PyTorch preferred).
Familiarity with inference and optimization toolchains such as quantization, distillation, and pruning (e.g., vLLM/SGLang, llama.cpp, MLX, CoreML).
PhD with research in efficient multimodal reasoning, LLM reasoning, model compression, or lightweight VLM architectures (preferred).
Experience training or post-training vision-language models end-to-end, including connector design and visual instruction tuning (preferred).
Experience deploying LLM or multimodal models on mobile or edge hardware using CoreML, MLX, or TensorRT-LLM (preferred).
Generative AI
Multimodal Foundation Models
Deep Learning
Python
PyTorch
JAX
Computer Vision
Data Pipelines
On-device Machine Learning
Data Curation
Vision-Language Models
Video-Language Models
Python
PyTorch
JAX
Deep Learning
Generative AI
Multimodal Foundation Models
Computer Vision
Data Pipelines
On-device Machine Learning
Data Curation
Vision-Language Models
Video-Language Models
Python
PyTorch
JAX
Deep Learning
Generative AI
Multimodal Foundation Models
Computer Vision
Data Pipelines
On-device Machine Learning
Data Curation
Vision-Language Models
Video-Language Models