Ciprian Bangu

Currently doing my M2 internship in the Neural Coding and Neuroengineering of Human Speech Functions lab at Pasteur Institute of Audition, supervised by Dr. Anne-Lise Giraud. The project consists of applying Convolutional Neural Nets and Vision Transformers to decode imagined speech from human magnetoencephalography (MEG) data.

Leveraged knowledge of embedding models to help design and develop a solution that integrated semantic information into comparable company search, to increase efficiency and accuracy in pre-deal valuation research. Stake-holders estimated this solution could reduce research time by about 20%. We came 2nd in the competition.

Researched data evaluation and visualization techniques for Large Language Model training and fine-tuning. • Ran Direct Preference Optimization fine-tuning experiments to determine the predictive power of data quality metric I developed. Experimental results suggest an increase in the model’s MMLU performance when trained on data organized by this metric. • Utilized the Bunka tools to create maps of textual datasets, and presented this work as a poster at the Hy’CHA 2024 conference.

Backpacked around the world with my partner for 7 months. Over the course of the trip, we visited 17 countries, 48 cities, and 3 continents. Afterwards, I began my Master's degree in Cognitive Science.

Responsible for all day-to-day operations for the Lost Instrument and Waiver of Probate insurance programs for the largest Canadian Stock Transfer Agents. Liaised with multiple stakeholders (Registered Holders & Personal Representatives, Underwriters, etc.) to facilitate the issuance of non-program bonds (i.e.,250kCAD +)

Logic Lab TA for PHL245: Modern Symbolic Logic

Logic Lab TA for PHL245: Modern Symbolic Logic