Senior AI/ML Engineer or Postdoctoral Researcher - LLM Infrastructure

The role focuses on scalable training and inference pipelines, token-efficiency and model selection strategies, distributed ML systems in production, and the evaluation of open models. It also includes optimizing latency, cost and performance trade-offs, supporting parameter-efficient adaptation workflows, and ensuring reliable integration with orchestration layers in EOSC service environments.

Your responsibilities will include

• Designing and developing inference infrastructure in cloud environments, with continuous optimization for latency, throughput and cost efficiency
• Benchmarking models against task-specific and general-purpose evaluation criteria and deriving data-driven recommendations on model-task fit
• Applying quantization, distillation, parameter-efficient fine-tuning and other optimization techniques to improve efficiency, performance and cost-effectiveness
• Building an orchestration layer that integrates seamlessly with the broader agent framework and supports emerging AI protocols such as MCP, A2A and ACP
• Embedding observability from the outset using tools such as OpenTelemetry and Grafana to make model performance, cost and reliability transparent, measurable and actionable

We are seeking a Senior AI / ML Engineer interested in building, optimizing, and operating general-purpose AI assistants at scale within the European Open Science Cloud (EOSC) ecosystem. The successful candidate will have a strong background in machine learning and artificial intelligence, sound mathematical understanding, and practical experience with LLM inference optimization in cloud environments., Your profile

• Completed scientific university degree (Master's degree or equivalent) in computer science, software engineering, artificial intelligence, data science or a related field
• Proven experience with LLM inference optimization in cloud environments
• Practical experience with fine-tuning, quantization, parameter-efficient fine-tuning (PEFT) and deployment of large language models in production or production-like settings
• Experience with model profiling, benchmarking and establishing performance baselines
• Experience in designing, developing or operating scalable training and inference pipelines within distributed AI/ML systems
• Knowledge of multi-agent systems, agentic workflows or orchestration frameworks, such as LangGraph or AutoGen
• Experience working with open-source LLMs, such as Llama or Mistral
• Very good programming skills in Python and experience in developing reproducible code, e.g. using PyTorch or comparable frameworks
• Good knowledge of cloud-native technologies, in particular Docker and Kubernetes
• Very good written and spoken English skills

Desirable additions to your profile

• Completed PhD in a relevant field
• Experience with agentic design patterns, such as reflection, ranking, exploration/discovery or human-in-the-loop feedback loops
• Experience handling scientific datasets and applying FAIR principles to data, model outputs or research workflows
• Strong written communication skills and experience in interdisciplinary collaboration, including contributions to scientific publications
• Familiarity with the Rust programming language, In the case of foreign university degrees, it is mandatory to submit a Statement of Comparability from the Central Office for Foreign Education (ZAB) for the final assessment of the hiring requirements during the hiring process (link for more information). A fee will be charged for this. This obligation does not apply if the foreign university degree has already been assessed by an expert and is listed in the ANABIN database of the ZAB as an equivalent university degree (link to ANABIN database)., + Natural sciences and mathematics

• Computer science
• Mathematics
• Physics

Level of education: Completed scientific university degree (Master's degree or equivalent) in computer science, software engineering, artificial intelligence, data scienc

Working language and expected level:

• English (Very good command of the language)

The position is initially limited to the project duration of three years with envisioned extension. An extension is possible. The regular weekly working hours are 39.8 hours (full-time). The position is generally suitable for part-time work. The position is generally expected to be carried out on site in Hanover. Mobile work is possible to a certain extent, subject to operational requirements. The remuneration is based on pay scale group 13 of the collective agreement for the public service of the German states (TV-L).

What we offer Our mission is to keep rethinking and innovating the provision and use of research data and information. In TIB's Research and Development Department you have the opportunity to advance your career in a sizable, dynamic and excellent environment. We provide a scientifically and intellectually inspiring environment with an entrepreneurial mindset, embedded in a leading technical university and one of the largest technical information centers being part of the Leibniz Association. TIB collaborates closely with the L3S Research Center at Leibniz University Hannover, one of the world's leading research institutes in the field of Web & Data Science, within the Leibniz Joint Lab Data Science & Open Knowledge. Last but not least, we attach great importance to an open, creative and fun work atmosphere.

Furthermore, we offer

• A job in the public service oriented towards the common good on the basis of the collective agreement for the public service of the German states (TV-L) with a salary in pay scale group 13 TV-L
• A flexible workplace in terms of time and space with offers to reconcile work and family life, such as mobile and remote work options as well as flexible working time models (flexitime)
• A special annual payment at the end of the year and 30 days of vacation per year with a five-day working week as well as additional days off on Christmas Eve (December 24th) and New Year's Eve (December 31st)
• A modern workplace in a central location of Hannover with a collegial, attractive and versatile working environment
• An employer with a wide range of internal and external further education and training measures, workplace health promotion and a supplementary pension scheme for the public sector (VBL)
• Employee discount in the canteens of the Studentenwerk Hannover as well as the possibility to use the various offers of the University Sports Hannover at a discount
• Independent and future-oriented activities that offer variety and room for personal development
• Funding for necessary equipment, conference and research visit travel
• Work in the context of a national, European or international research and innovation project
• A portfolio of technology components to build on, including ORKG, OpenResearch.org, TIB AV-Portal, DBpedia.org and other

Die TIB – Leibniz-Informationszentrum Technik und Naturwissenschaften und Universitätsbibliothek versorgt als Deutsche Zentrale Fachbibliothek für Technik sowie Architektur, Chemie, Informatik, Mathematik und Physik Wissenschaft, Forschung und Wirtschaft mit Literatur und Information. Sie hat die Aufgabe, das verzeichnete Wissen zu erhalten und aktuelle Informationen unabhängig von Ort und Zeit heute und in Zukunft bereitzustellen. Die TIB engagiert sich für Open Access und unterstützt damit den unbeschränkten und kostenlosen Zugang zu wissenschaftlicher Information. In ihrer Funktion als Universitätsbibliothek stellt sie die Informationsversorgung aller Fakultäten der Leibniz Universität Hannover sicher.

Die TIB baut ihre Rolle als deutsches Informationszentrum für die Digitalisierung von Wissenschaft und Technik stetig weiter aus. Als forschende Bibliothek betreibt sie angewandte Forschung und Entwicklung, um neue Dienstleistungen zu generieren und bestehende zu optimieren. Die Schwerpunkte liegen auf Data Science & Digital Libraries, nicht-textuellen Materialien, Open Knowledge, Open Science und Visual Analytics.

Für Fach- und Forschungscommunities stellt die TIB unter www.tib.eu wissenschaftliche Inhalte und digitale Dienste bereit, um die verschiedenen Phasen des wissenschaftlichen Arbeitens zu unterstützen. Über ihr Recherche- und Bestellportal bietet die Bibliothek Zugriff auf ihren exzellenten Bestand an grundlegender und hoch spezialisierter technisch-naturwissenschaftlicher Fachinformation. Hierzu zählen auch Wissensobjekte wie 3D-Modelle, Forschungsdaten und audiovisuelle Medien. Im AV-Portal der TIB können wissenschaftliche Videos aus Technik und Naturwissenschaften zielgenau durchsucht werden. Durch die Vergabe von DOI-Namen (Digital Object Identifier) sichert die TIB die Qualität sowie die dauerhafte Verfügbarkeit von 
Forschungsdaten.

Die TIB ist eine Stiftung öffentlichen Rechts des Landes Niedersachsen. Sie ist Mitglied der Leibniz-Gemeinschaft.