Deep Learning Scientist - Foundation Models

About TranslatedTranslated is on a mission to allow everyone to understand and be understood, in their own language. We are a technology-powered professional translation provider. We partner with over 200,000 professional translators worldwide, in 200 languages. Our 310,000 clients range from private individuals who need their CV translated to some of the world's largest companies, such as Uber and Airbnb.Our progress is largely powered by our ability to leverage scientific progress and realize the best synergy between humans and machines. We invest heavily in R&D, including Large Language Models, Machine Translation, expressive speech synthesis, and privacy-preserving training. We operate as a science-driven company, where scientific innovations quickly make their way into production and have a measurable impact on our products and operations.We are looking for a Deep Learning Scientist to join our Foundation Models team, working on the development of large-scale textual language models from scratch.The ideal candidate has a strong interest in Large Language Models, large-scale deep learning, and experimental research, and is excited about pushing the state of the art in multilingual language modeling.The Project: Foundation ModelsTranslated is building its own generation of multilingual Foundation Models, with the ambition of advancing the state of the art in open Large Language Models, particularly in multilingual capabilities.Our work covers the complete LLM pre-training lifecycle: from data selection and curation to model architecture, scaling experiments, large-scale distributed training, evaluation, and continued training.A central research question for the team is how to build models that perform strongly not only in English, but across a broad range of languages, including languages that are traditionally underrepresented in large-scale pre-training datasets.This involves research across areas such as:large-scale language model pre-trainingmultilingual language modelingdata curation, filtering, scoring, and mixture designscaling laws and training dynamicsmodel architecture and optimizationdistributed and multi-GPU trainingevaluation and benchmarking of Large Language Modelscontinued pre-training and mid-training strategiesYou will work in the Foundation Models team, a multidisciplinary group of scientists and engineers responsible for designing, training, and evaluating Translated’s next generation of Large Language Models.The team works closely with Translated's AI Research and Engineering organizations, combining experimental research with the engineering required to train models at scale on large GPU clusters and HPC infrastructure.Responsibilitiesdesign and conduct research on Large Language Model pre-trainingdesign experiments, implement them in code, run them at scale, and analyze their resultsinvestigate model architectures, optimization strategies, training dynamics, and scaling behaviorresearch and develop multilingual training strategieswork on data selection, quality, composition, and data mixture experiments for LLM pre-trainingevaluate models across multilingual and general-purpose benchmarksmonitor and benchmark the state of the art in Large Language Modelsrun experiments on large-scale GPU and HPC infrastructuretranslate research hypotheses into measurable experiments and actionable decisions for large-scale trainingRequirements3+ years of research/industry experience in a relevant area of Deep Learning, Machine Learning, Natural Language Processing, or Large Language Modelsstrong understanding of modern deep learning and Transformer-based language modelsexcellent programming skills in Pythonexperience designing and running machine learning experimentsfamiliarity with GPU-based training environments and Unix/Linux systemsability to analyze experimental results and make research decisions based on empirical evidenceinterest in large-scale experimental research and language model pre-trainingability to follow, understand, and reproduce recent scientific literatureexcellent written and spoken Englishability to work effectively with both researchers and engineersBonus points if...you have direct experience pre-training or continuing the pre-training of Large Language Modelsyou have experience with distributed and multi-GPU trainingyou have worked with large-scale training frameworks such as Megatron Bridgeyou have experience optimizing GPU utilization, throughput, memory consumption, or large-scale training stabilityyou have worked on multilingual NLP or multilingual language modelsyou have experience with large-scale dataset curation, filtering, deduplication, or data mixture designyou have experience with LLM evaluation and benchmarkingyou have experience working with HPC environmentsOur OfferDepending on expertise, the offered salary typically ranges between €40.000,00 and €55.000,00. Compensation generally grows quickly as experience leads to greater contributions. Exceptionally strong candidates may be offered salaries above this range to acknowledge their higher potential impact.We offer a flexible smart-working policy, while requiring regular presence at our Rome headquarters.HeadquarterTranslated is hosted at Pi Campus, a working environment immersed in nature where six luxury villas in Rome, Italy, have been converted into functional offices designed to foster talent growth. Pi Campus is also a venture firm created by Translated to reinvest part of its profits into promising AI startups.Benefits and PerksAt Translated, we see our people as athletes, and Pi Campus as the place where they can reach their full potential. This vibrant environment fosters talent aggregation and continuous growth of mind, body, and spirit.We nurture and support the team's personal and professional development every day through growth-oriented initiatives such as personalized learning paths and networking events with industry leaders; health-oriented initiatives including massages, sauna, workouts in our gym and swimming pool; as well as relaxation rooms, open meeting rooms, a cafeteria, and fully equipped kitchens in every villa.In case of need, every team member has access to psychological, legal, and financial support.Learn more about our company: StatementAt Translated, we proudly embrace and celebrate each individual's unique qualities, regardless of race, sexual orientation, gender identity, or any other differences. We recognize that diverse perspectives empower us to overcome challenges, foster innovation, and drive excellence.As an inclusive and equal-opportunity employer, we are committed to cultivating an environment where everyone feels welcome, valued, and supported to achieve their full potential.SummaryLocation: Roma, RM, ItalyType: Full TimeExperience: Mid LevelDepartment: Engineering