LLM-Assisted Query Understanding and Retrieval Internship

Are you interested in figuring out how Large Language Models can be used to improve user experience in Maps Search? Do you enjoy experimenting with LLM fine-tuning and semantic search and measuring the impact of experimental models on search quality metrics and actual user interactions? Then we would love to hear from you!

What you'll do

• Fine tune Large Language Models for the query segmentation task in maps search domain (geocoding).
• Explore different approaches of organizing search indexes, which can leverage query segmentation information to improve search latency and relevance.
• Collaborate with engineers to integrate query segmentation models into production TomTom Search API.
• Explore ways to index map data for fast retrieval, alternative to traditional inverted index approach, be it prefix index or embeddings.
• Benchmark integrated solutions on search quality and latency metrics.
• Collaborate with data scientists and engineers to analyze the impact of query segmentation on TomTom’s Search APIs offline and online search quality metrics.
• Contribute to reusable components, pipelines and documentation that can be used beyond the internship.

What you'll need

• You are enrolled as a full-time student for the entire duration of the internship.
• You have EU citizenship or are enrolled in a Dutch university (due to work permit regulations).
• You have a background in Computer Science, AI, Data Science, Information Retrieval or related field.
• You are available to start in February or March.
• Strong programming skills, preferably in Python. Experience in Java or Scala is a plus.
• Basic knowledge of machine learning concepts, information retrieval concepts and experimentation.
• Basic knowledge of search libraries and platforms, like Lucene or OpenSearch/ElasticSearch.
• Interest in Search and Machine Learning domains.
• Comfortable working with data, running experiments, and analyzing results.
• Familiarity with LLMs (e.g. BERT etc.) is a plus, but not required.
• A curious, exploratory mindset and willingness to explore different technical approaches.