A new approach to AI translation

https://seas.harvard.edu/news/camilo-brown-pinillas-senior-project-new-approach-ai-translation

Publish Date: 2026-04-06 07:16:00

Source Domain: seas.harvard.edu

Project Overview: Fourth-year applied mathematics students at Harvard’s SEAS can write senior theses that employ mathematical, statistical, or computational modeling to explore phenomena with a focus on causation and mechanism.
Specific Thesis Topic: Camilo Brown-Pinilla’s senior thesis investigates Natural Language Processing from a Geometric Deep Learning perspective, treating translation between languages as a symmetry and exploring how this notion could enhance AI translation models.
Project Genesis: The idea originated from a junior-year class on Geometric Machine Learning, combined with his background in Natural Language Processing, and a desire to bridge these two fields.
Timeline: Initiated in April 2025, the project involved extensive literature review and problem formulation in the summer of 2025, followed by the implementation and experimentation from August 2025 to March 2026.
Challenges: The largest challenge was formulating an actionable research question by bridging the gap between Geometric Deep Learning and Natural Language Processing. Additionally, the engineering pipeline required optimizing and orchestrating large models and datasets.
Enjoyable Aspects: Camilo enjoyed the deep exploration of a fascinating problem, appreciating the process of integrating various techniques and seeing how they all connected together.
Skills and Learning: The project taught Camilo how to think like a scientist, fostering curiosity, and translating it into tangible results, gaining numerous new skills in the process.