Marcel Mateos Salles
Hey! I'm Marcel, a recent grad from Brown University with a degree in Computer Science and Economics. I'm a backend engineer and ML/AI researcher with a passion for intersecting software engineering with cutting-edge research in ML/AI. My current research interests lie in LLMs, model interpretability, and self-supervised learning. I'm a member of the GalilAI Group with Prof. Randall Balestriero and an incoming Software Engineer at Pinterest!
Experience
Software Engineer
Incoming summer 2026
Incoming July 2026Head Teaching Assistant
CSCI1460: Computational Linguistics
Brown University
Dec 2025 - May 2026Software Engineer Intern
Backend - Observability Team
Developed GenAI tools (such as agents, MCP servers, anomaly detection) for all engineers at Pinterest.
Summer 2025Teaching Assistant
CSCI 1470: Deep Learning
Brown University
Dec 2024 - May 2025Researcher
GalilAI Group @ Brown University
Researching spurious correlation in language models. Check out our new paper! Papers in workshops at NeurIPS 2025 (poster) and ICLR 2026 (oral).
Oct 2024 - PresentSoftware Engineer Intern
Dexcom
Developed architecture for a new mobile app and designed a novel algorithm to target 700k new stakeholders
Jun 2024 - Aug 2024Teaching Assistant
Econ 1000: Using Big Data to Solve Social and Economic Problems
Brown University
Sep 2023 - Dec 2023Researcher
SAVE Lab @ UTSA
Updated tensorflow models / data pipelines
May 2023 - Sep 2023Bachelor of Science in Computer Science-Economics, Magna Cum Laude, Honors in Computer Science-Economics
Brown University
GPA: 4.0/4.0
Honors Thesis: When Efficiency Enables Shortcuts: Studying Spurious Correlations Under LoRA Finetuning
Sep 2022 - May 2026Programming Languages
Frameworks/Libraries
Recent Projects
When Efficiency Enables Shortcuts: Studying Spurious Correlations Under LoRA Finetuning
Honors thesis investigating how LoRA finetuning can introduce and amplify spurious correlations in language models.
Reddit Depression Symptom Detection
A computational linguistics project that uses Reddit data to detect symptoms of depression.
SpotiDuo
A music-based language learning app featuring support for over 150 languages, user inputs, and a recommendation algorithm to continue learning
Little Label Learners
A novel semisupervised learning technique based on biological development.
SpacePals
Hack@Brown 2024 project aimed at entertaining users through a pokemon-like game of space exploration.
Campfire Stories
Hack@Brown 2023 project aimed at improving camping safety through storytelling.
Get in Touch
Think I'd be a good fit for your team? I'm always interested in hearing about new projects, opportunities, and challenges!
Email Me!