AI Research Engineer (Model Compression & Quantization) - 100% Remote Worldwide

4 months ago

Remote, WorldwideSenior

Responsibilities

Design and deploy high-throughput, low-latency model-serving architectures with optimized memory usage across mobile, edge, and other resource-constrained environments.
Establish performance targets and monitor latency, throughput, memory consumption, error rates, and other production inference metrics.
Build and run controlled inference tests in simulated and live production environments, documenting results against established benchmarks.
Prepare test datasets and simulation scenarios for evaluating model performance on low-resource devices.
Analyze serving-pipeline efficiency and resolve bottlenecks involving batching, network delays, processing, and memory usage.
Integrate optimized serving and inference frameworks into production pipelines for edge and on-device applications.
Apply distributed inference techniques, including tensor parallelism, pipeline parallelism, and expert parallelism, to support large models on GPU clusters.
Iterate on model-serving and inference optimizations to improve real-world performance, reliability, scalability, and memory efficiency.

Bachelor's degree in Computer Science or a related field; a PhD in NLP, Machine Learning, or a related field is preferred.
Solid AI R&D track record with strong publications in A* conferences preferred.
Knowledge of Metal Shading Language and ability to write custom compute shaders from scratch.
Proven experience with low-level kernel and inference optimization on mobile devices, including measurable improvements in latency, throughput, and memory footprint.
Deep understanding of modern model-serving architectures, inference engines, inference optimization, and efficient memory management.
Strong expertise writing GPU kernels for mobile devices such as smartphones.
Practical experience developing and deploying end-to-end inference pipelines on resource-constrained devices and edge platforms.
Ability to apply empirical research, design evaluation frameworks, and address latency, computational, and memory constraints.
Understanding of Diffusion Models and Vision Transformers, along with pruning, quantization, FlashAttention, KV caching, and speculative decoding such as Eagle.

100% remote work worldwide.
Opportunity to work on advanced AI serving and inference systems within a global fintech and digital-asset company.
Collaboration with cross-functional teams on production edge, on-device, and GPU-cluster applications.

201-500 employees