Senior Machine Learning Engineer - Recommendations (Core)
SoundCloud1 day ago
Berlin, Germany or London, United KingdomSenior
Responsibilities
- Design, train, and ship recommendation models for candidate generation, user and item modeling, ranking, search, and content understanding.
- Lead the migration of key models from classical approaches such as word2vec-style embeddings to PyTorch models trained and served on GPUs.
- Adapt, fine-tune, and evaluate open-source embedding and ranking models for SoundCloud’s catalog and use cases.
- Own offline metrics, A/B experiments, and the validation that offline improvements translate to online gains.
- Collaborate with ML, data, and backend engineers to productionize models, maintain their health, and establish modeling standards.
Requirements
- At least 3 years of experience building production ML systems and end-to-end ownership of a ranking, retrieval, or recommendation model.
- Strong deep learning skills in PyTorch or similar frameworks and knowledge of recommendation and retrieval concepts including embeddings, two-tower models, learning to rank, and multi-objective ranking.
- Experience evaluating and experimenting with ranking systems, including handling biases in implicit feedback such as clicks and plays.
- Strong Python skills, large-scale data processing experience with tools such as Spark or BigQuery, and experience deploying models to a cloud platform, preferably GCP.
- M.Sc. in a quantitative field or equivalent experience.
- Preferred qualifications include transformer-based embedding or ranking model fine-tuning or serving, LLM or foundation-model applications, audio or music ML, multimodal representations, contextual bandits, reinforcement learning, causal methods, relevant publications, ANN or vector search, and GPU training or serving at scale.
Tech Stack
Categories
About SoundCloud
SoundCloud builds a music streaming and audio platform for artists and fans, plus tools for creators to distribute, monetize, and grow. Revenue comes from listener subscriptions and ads, and from paid creator services such as distribution and promotion. Founded in 2007 and headquartered in Berlin, the company operates globally as a privately held business.