Rocco Di Tella

rocco.di_tella at cfa.harvard.edu

I use tools Machine Learning and Statistics for Neuroscience and Astrophysics.

My current projects are in the context of AstroAI, Harvard & Smithsonian Center for Astrophysics.

MSc in Data Science from Universidad de Buenos Aires.

Currently based in Boston.

Ongoing Projects

Multi Modal Galaxies | AstroAI, Harvard Center for Astrophysics

Experiments on Transformer-based Flow Matching to generate multimodal joint distributions.
Honest LLMs through introspection | AI Safety Argentina \& ICC

Statistically sound measures of LLM confidence with probes.

Visual Attention | Kreiman Lab, Harvard Medical School

Do Neural Networks implicitly model Human visual attention?
Active Learning for more efficient LLM rankings | ICC

Leveraging statistical active learning to make more efficient LLM comparisons
Fast non-parametric Bayesian algorithm for more efficient online experiments | Nueroscience Lab, UTDT

A very fast algorithm that enables experiments to optimize their tasks on the fly

Course coming up this August: Machine Learning for Astrophysics | UBA
Three separate graduate courses on: Statistics, Bayesian Statistics, Causal Inference | UBA
One very hands on course on Machine Learning | UdeSA

(In spanish)