Internship - Machine Learning Research Engineer

Overview

Perplexity is offering a Machine Learning Research Engineer internship in its Search team at the Berlin office. The internship programme runs for 12 to 24 weeks, full time and in person. The work centres on pushing search quality forward through models, data, tools or any other available means.

What you will do

Interns train and optimise large-scale deep learning models with frameworks such as PyTorch, using distributed training approaches including PyTorch Distributed, DeepSpeed and FSDP together with hardware acceleration, with a focus on retrieval and ranking models. They also conduct research in representation learning, including contrastive learning, multilingual work, evaluation and multimodal modeling for search and retrieval, and build and optimise retrieval-augmented generation pipelines for grounding and answer generation.

Requirements

Applicants should understand search and retrieval systems, including quality evaluation principles and metrics, and have strong proficiency with PyTorch, including distributed training techniques and performance optimisation for large models. An interest in representation learning is expected, covering contrastive learning, dense and sparse vector representations, representation fusion, cross-lingual alignment, training data optimisation and robust evaluation. A publication record at AI and machine learning conferences or workshops such as NeurIPS, ICML, ICLR, ACL, EMNLP or SIGIR is listed among the qualifications.