Scholar
Ernests Lavrinovics
Google Scholar ID: SR-WcVYAAAAJ
PhD Fellow at Aalborg University
Citations & Impact
Citations
66
H-index
3
i10-index
2
Publications
7
Co-authors
7
list available
Publications
Resume (English only)
Academic Achievements
- Focused on LLM factuality, low-resource constraints, and explainable language modeling. Has experience in various areas such as image coloration, novel view synthesis, and speech enhancement using diffusion-based generative models.
Research Experience
- Worked for nearly 3 years at Fraunhofer IIS institute as part of an audio research team developing a proposal for the MPEG-I standard. His role included planning and developing test content in XR, conducting listening tests, developing a common audio evaluation platform, and developing CI/CD pipelines with GitLab. For almost the past 2 years, while studying at Aalborg University, he has also been working as an Android app developer for a local startup, contributing to maintaining and developing new features for the Android app. Additionally, he worked as a Student Research Assistant developing a prototype of a physiotherapy AI-based evaluation system and as a Teaching Assistant for an introductory Machine Learning course at Aalborg University, Copenhagen.
Education
- PhD Fellow at Aalborg University Copenhagen since September 2024, focusing on LLM factuality, primarily looking into hallucination phenomena and methods of mitigating it with the use of knowledge-graphs. During his master's studies, he worked on projects related to AI transparency and adversarial attacks, Question-Answering, Part-of-Speech tagging, and parameter-efficient (adapter-based) Machine Translation within NLP.
Background
- Research interests include Natural Language Processing (NLP) and Machine Learning (ML), particularly LLM factuality, multilinguality, knowledge graphs, and parameter-efficient fine-tuning. Interested in the ethical implications of AI and its role in society.
Miscellany
- Interested in developing cross-modal AI systems that can learn from multiple sources of data, such as text and audio or text and images.
