AI Research Engineer (Model Compression & Quantization)

4 months ago

Remote, WorldwideSenior

Responsibilities

Apply low-bit quantization to LLMs, VLMs, and other multimodal generative AI models while maintaining accuracy and output quality.
Use knowledge distillation to transfer capabilities from large teacher models to smaller multimodal student models.
Implement pruning methods to remove redundant parameters and attention heads without sacrificing task performance.
Analyze model-size, latency, memory, throughput, and accuracy trade-offs and propose improvements based on empirical results.
Research and apply mixed-precision quantization, adaptive pruning schedules, intermediate-feature distillation, and other advanced compression methods.
Build compression pipelines, establish performance and fidelity metrics, and address production inference bottlenecks on edge devices.
Stay current with model-compression research for multimodal and generative architectures.
Document methodologies, experiments, and results to support reproducibility and collaboration.
Author technical papers and publish findings at conferences including NeurIPS, ICML, ICLR, CVPR, ACL, and AAAI.

Degree in Computer Science or a related field.
Strong AI research and development background with a track record of publications at leading conferences; a PhD in NLP, Machine Learning, or a related field is preferred.
Hands-on experience with PyTorch or an equivalent deep learning framework.
Experience with both Quantization-Aware Training and Post-Training Quantization.
Research and hands-on experience with knowledge distillation for compressing large models.
Research and hands-on experience with model pruning for compressing large models.
Strong understanding of neural network architectures and training, including transformers, LLMs, VLMs, backpropagation, optimization, and fine-tuning.
Familiarity with C++ is a plus, particularly for low-level quantization kernels or inference optimization.

201-500 employees