Silas Alberti

I am a PhD student in machine learning at Stanford University. I am co-advised by Stefano Ermon and Chris Ré and I am broadly interested in foundation models and theory. I am affiliated with Stanford AI Lab and Center for Research on Foundation Models. My undergrad was in Mathematics at LMU Munich, where I was a research assistant at Prof. Gitta Kutyniok’s chair for Mathematical Foundations of AI and helped coordinate the DFG-funded priority program “Theoretical Foundations of Deep Learning”. Moreover, at UC Berkeley I was advised by Prof. Song Mei and conducted research with Prof. Bruno Olshausen from the Redwood Center through Machine Learning at Berkeley.

Last summer, I was a quantitative trading intern at Jane Street and for the past 2 years I was working as a part-time software engineer at the startup Fulfin. Currently building Intern Chat.

Education

Stanford University, 2022 - present

Ph.D. Student, Electrical Engineering
University of California, Berkeley, 2021

Visiting Student
LMU Munich, 2019 - 2022

B.Sc. in Mathematics & Computer Science

Research

Deep Learning Theory

Studying the effects of inductive biases on generalization of neural networks and kernel methods. Advised by Prof. Song Mei at UC Berkeley.

Local Capsule Hierarchies

New architecture for learning deformable part-whole hierarchies. Collaboration with Domas Buracas and Nitish Dashora from Machine Learning at Berkeley, advised by Prof. Bruno Olshausen.

Mathematical Foundations of Artificial Intelligence

Developing a new benchmark and studying new architectures for letting Graph Neural Networks detect directionality. Advised by Prof. Gitta Kutyniok and Dr. Ron Levie.

Projects

Cryptoranker

Animated visualization of the market caps of the top 20 cryptos over time. Frontend in Javascript that can be fed with any CSV data. Crypto data scraper in Python. See on cryptoranker.alberti.xyz. Code on Github.

Comprehendum

GPT-3 🔁 Slack. A convenient Slack command that answers natural questions about a set of documents, e.g. legal & compliance, using GPT-3. Code on Github.

Fractal Explorer

Infinite fractal zoom implemented in Java. Various fractals implemented, e.g. Mandelbrot set, Julia sets and more. Customizable color schemes. Code on Github.