Nico Previtali

• Data Scientist und AI Engineer mit einem Bachelor-Abschluss in Artificial Intelligence & Machine Learning von der Hochschule Luzern.
• Umfassende Erfahrung in der Anwendung von AI/ML-Techniken in verschiedenen Bereichen und Domänen, einschließlich Recommender-Systeme, Bild- und Klangverarbeitung sowie Natural Language Processing (NLP). 
• Bachelorarbeit zur Analyse großer Sprachmodelle in NLP. 
• Vertraut mit Python, PyTorch, SQL, Docker, Git, … 
   

• Bachelor Thesis: Can Large Language Models Follow Discussions? , Rotkreuz 
  Developed linguistic probing tasks and evaluated conversational understanding of various large language models (LLMs) such as GPT, BERT, and ELEKTRA based on user data scraped from the Kialo debate platform. Authored a scientific report in English comparing the performance of these models. 
• Movie Recommender System in MySQL Feb 2023 - Jun 2023 HSLU, Rotkreuz 
  Developed and implemented a recommender system using minimal user data and based on ratings from MovieLens and IMDb. Conducted specific data analyses to enhance movie selection personalization. Visualized the query results by using the tool Metabase to improve user experience and business intelligence. 
• Neural Style Transfer with PyTorch Feb 2023 - Jun 2023 HSLU, Rotkreuz 
  Fine-tuned a CNN (VGG19) for style transfer on images, experimenting with various parameters for style and content loss weighting. Applied to images and videos, including feature map visualizations 
• Timbre Transfer with AI from Midi Data Sep 2022 - Dec 2022 HSLU, Rotkreuz 
  Prepared the NSYNTH dataset in TensorFlow Record format for this application and trained a neural network using Google Magenta’s DDSP library to effectively transfer timbral characteristics between different musical instruments. 
• Protein Stability Prediction (Kaggle Challenge) Sep 2022 - Dec 2022 HSLU, Rotkreuz 
  Collaborated, fine-tuned, and evaluated CNN and XGBoost (decision tree) models on 3D protein data from AlphaFold to predict thermal stability after mutations for the Kaggle Challenge Novozymes Enzyme Stability Prediction. 
• Optimization of Language Models for Multilinguality (Kaggle Challenge) Sep 2022 - Dec 2022 HSLU, Rotkreuz Collaborated, fine-tuned, and evaluated XLM-RoBERTa and LSTM models on multilingual natural language inference in PyTorch, followed by comprehensive performance and error analysis for the Kaggle Challenge Contradictory, My Dear Watson. 
• Recommender-System für Jointcreate Sep 2021 - Jun 2022 Jointcreate, Rotkreuz 
  Collaborated on the conceptual development and implementation of a recommender system in Python to predict user preferences based on partially self-annotated labels for the open innovation platform jointcreate.com . 
• Data Visualization: Chess Tournament Big Data Sep 2021 - Dec 2021 HSLU, Rotkreuz 
  Designed and implemented an interactive sunburst visualization using D3.js in the Observable framework to represent and analyze opening moves from a chess tournament dataset from Kaggle. 
• Web Development in HTML, CSS & PHP Sep 2020 - Dec 2020 HSLU, Rotkreuz 
  Developed a responsive website using HTML, CSS, JavaScript (Canvas), PHP for server-side scripting and form validation, and W3.CSS for layout design.

Sep 2020 - Jan 2024 Bachelor of Science in Artificial Intelligence & Machine Learning Hochschule Luzern / FHZ, Rotkreuz

Aug 2018 - Jul 2020 Application Development WISS, Zürich Way-Up apprenticeship 

2017 - 2018 Psychology and Mathematics University of Zurich, 

2013 - 2015 Zürich Biology with Laboratory Internship ETH, Zürich Laboratory Internship 2x per week 

2007 - 2013 Gymnasium, Specialization: Latin with English Immersion (Cambridge English C2 Proficiency)
Kantonsschule Wiedikon, Zürich