Google logo

Research Scientist, ML Efficiency, Google Research

Google

SingaporeFull-timeOn-site

Posted 24 Jul 2026

About this role

The role focuses on advancing algorithms and architectures to improve the computational efficiency and serving of large-scale generative AI models. Responsibilities include optimizing latency, memory bandwidth, and deployment pipelines through collaboration with hardware and software teams. Candidates must hold a PhD in Computer Science or a related field and have a record of scientific publications in top-tier AI conferences. Preferred qualifications include expertise in transformer internals and the ability to lead research from abstraction to production.

What they're looking for

Machine LearningDeep LearningTransformer ArchitectureModel CompressionKnowledge DistillationQuantizationInference AccelerationLarge Language Models

Frequently asked questions

What does a Research Scientist, ML Efficiency, Google Research at Google do?

The role focuses on advancing algorithms and architectures to improve the computational efficiency and serving of large-scale generative AI models. Responsibilities include optimizing latency, memory bandwidth, and deployment pipelines through collaboration with hardware and software teams. Candidat…

What skills does this Research Scientist, ML Efficiency, Google Research role need?

Key skills for this role include Machine Learning, Deep Learning, Transformer Architecture, Model Compression, Knowledge Distillation, Quantization.

How much does a Research Scientist, ML Efficiency, Google Research at Google pay?

The employer did not list a salary for this role. Most similar Singapore roles publish their band on the job page.

Is this Research Scientist, ML Efficiency, Google Research role remote, hybrid, or on-site?

This role is on-site, based in Singapore.

How do I apply for this Research Scientist, ML Efficiency, Google Research role?

You can apply directly on Google's careers page. ApplyLah can tailor your résumé and cover letter to this exact role in seconds first.