PhD or Postdoc Research Position: Web-Search Enabled LLMs for a Circular Economy M/V/X

Description de l'entreprise

The KU Leuven Life Cycle Engineering (LCE) research group in the Department of Mechanical Engineering and the KU Leuven Institute for Sustainable Metals and Minerals (SIM2) has developed strong, unique expertise in reuse, repair, repurposing, remanufacturing, and recycling, in close cooperation with its industrial partners. To support these research activities, KU Leuven established the Re- and Demanufacturing Lab in Heverlee, where the multidisciplinary team develops state-of-the-art automation, spectroscopic, and computer vision equipment for material characterization, human-robot cooperative disassembly and sorting, product identification, and state evaluation. Within the REINFUSE project, the ambition is also to frame these previously developed tools within the broader ecosystem for acquiring, handling, and reselling used electronics and industrial equipment. Therefore, the REINFUSE project aims to develop the following digital infrastructure that complements and enhances the lab's research and development activities:
- Advanced product (re-)identification,
- AI-driven product research and data acquisition, and
- Lifecycle product traceability

Description de la fonction

The KU Leuven Life Cycle Engineering research group is expanding its activities at the intersection of re- and demanufacturing and advanced AI. Building on our infrastructure for product identification, computer vision, and lifecycle assessment, we will launch the REINFUSE project in 2026, a four-year Flanders Make SBO project.

REINFUSE aims to build the next generation of AI-augmented tools for product (re-)identification, information acquisition, product-data enrichment, valuation, and lifecycle decision support. Diverse inputs, such as product images, OCR text, label images, manuals, and web documents, are to be structured and validated within a product data backbone that supports reuse, repair, refurbishing, remanufacturing, and recycling. The project targets TRL 5 demonstration, combining fundamental research with robust, reusable software demonstrators for industrial validation with 10 Flemish companies active in production, refurbishing, remanufacturing, and online auctions.

We are seeking a PhD or Postdoctoral researcher to co-develop web-search-enabled LLM-based enrichment pipelines, schemas, and data-structure logic that form the core of the REINFUSE product-information backbone. This role combines:

• Research: designing, evaluating, and refining methods for multimodal product-data extraction, enrichment, and validation.
• Engineering: building and validating reliable, scalable, and maintainable codebases and pipelines that feed into REINFUSE demonstrators in close cooperation with industrial 

The core responsibilities include:

• (Co)design and develop the database structures that form the backbone of the product-information layer, ensuring that identifiers, attributes, and derived data are stored, versioned, traced, and validated over time.
• Implement and evaluate the data-enrichment pipelines that use web-search-enabled LLMs to extract model- and device-level properties from documents, manuals, and online sources. You will explore versioning strategies, provenance tracking, conflict resolution, and reliability scoring to build a continuously improving product knowledge base.
• Progressively integrate additional input modalities. Initially, image- and OCR-based identifiers are added to the structured database and enrichment logic; later, multimodal LLM-inferred attributes will be added. You will ensure that these inputs are reconciled, validated, and used to enhance product identification and valuation.
• Develop multimodal pipelines combining images, OCR text, free-text descriptions, and web-retrieved documents into structured product properties.
• Support the preparation and organisation of demonstrators and be involved in on-site user tests.
• Support the guidance of master's theses and job students supporting the validation cases and interface development.
• Present research results at (international) conferences and events.
• Assist in workshops, dissemination activities, and teaching tasks (for PhD researchers only at less than 10% of working time).

Profil

For both PhD and Postdoctoral candidates

• You hold a Master's degree obtained with cum laude or equivalent.
• You have strong programming skills, especially in Python, and are comfortable working with Git and API tooling, such as Postman.
• You have experience in machine learning, NLP/LLMs, multimodal systems, computer vision, or scraping.
• Having experience in data science with SQL or an ORM framework for designing and querying structured data is a strong asset.
• Experience developing software in a team environment and familiarity with collaborative development workflows (version control, code reviews, documentation) are a plus.
• You communicate effectively in English (oral and written); Dutch is an advantage, but not required.

Additional expectations for Postdoctoral candidates

• You hold a PhD in Engineering, AI, Data Science, or related fields with a high relevance to the project activities.
• You bring significant prior relevant experience, such as:

1. Designed complex data structures or schemas for real-world use,
2. Deployed and maintained web-searched LLM systems,
3. Built scalable data ingestion or enrichment pipelines

Compétences linguistiques

• Français (atout)

  • Comprendre : Expérimenté - (C1)
  • Écrire : Expérimenté - (C1)
  • Lire : Expérimenté - (C1)
  • Parler : Expérimenté - (C1)

Envie d'en apprendre davantage sur ce métier ou sur un métier proche de celui-ci ? Parcourez toutes les informations utiles sur Panorama des métiers. Panorama des métiers