Member of Technical Staff (Machine Learning Research Engineer)
About this role
Engineers who thrive on independent execution and complex problem-solving will find a compelling challenge advancing search infrastructure at Perplexity. The team builds next-generation search technologies, focusing heavily on retrieval, ranking, and answer generation to redefine how users find information globally. Daily work involves designing large-scale deep learning models, optimizing retrieval systems, and deploying models ranging from boosting algorithms to large language models.
This role exists to continually push search quality forward through novel models, data improvements, and robust tooling. Candidates who combine a strong academic or publication background with practical experience in large-scale search systems will integrate smoothly into this collaborative engineering environment based in Berlin.
What you will do
- Continuously improve search quality through advanced models, data improvements, and custom tooling.
- Design and construct core infrastructure components for the search platform and underlying model stack.
- Train and optimize large-scale deep learning models utilizing PyTorch and hardware acceleration techniques.
- Conduct cutting-edge research in representation learning spanning contrastive, multilingual, and multimodal methodologies.
- Deploy various models, including large language models and boosting algorithms, for high performance and scalability.
- Develop and refine retrieval-augmented generation pipelines to improve grounding and answer generation accuracy.
- Partner closely with data, artificial intelligence, infrastructure, and product groups to accelerate delivery timelines.
What you need
- Demonstrate a deep understanding of search and retrieval architectures, evaluation principles, and standard metrics.
- Bring a proven background working on large-scale search engines or recommender systems.
- Exhibit strong proficiency in PyTorch alongside practical expertise in distributed training and model optimization.
- Showcase specialized knowledge in representation learning, contrastive techniques, and embedding space alignment.
- Maintain a solid publication history in recognized artificial intelligence and machine learning conferences or workshops.
- Possess at least three to five years of professional experience focused on search, recommenders, or adjacent research domains.
Skills and tech stack
Questions candidates ask
Where is this job located?
This position is based in Berlin. Team members collaborate closely with data, artificial intelligence, infrastructure, and product groups locally and across the organization.
What frameworks and tools will I use?
Daily work centers around PyTorch, including distributed training tools such as PyTorch Distributed, DeepSpeed, and FSDP. Engineers also build and optimize retrieval-augmented generation pipelines and large-scale deep learning models.
What kind of background is required for this role?
Applicants need a minimum of three to five years of experience working on search, recommender systems, or closely related research areas. A strong publication record in major artificial intelligence or machine learning conferences is also expected.