Portfolio item number 1
Short description of portfolio item number 1
Posts by Collection
portfolio
Portfolio item number 2
Short description of portfolio item number 2
projects
Translating Tweets from Trumpese to Sanderese with Transformers and CycleGANs (2020)
Used Transformers to apply Cyclic Generative Adversarial Networks to the Natural Language Processing domain, attempting to transfer styles between tweets of different users
[Linguistics] Non-adult Behavior of Children’s Quantification in Logical Deduction Outside of the Language Domain (2020)
Proposed a psycholinguistic experiment to evaluate whether exhaustive pairing, a non-adult judgement common in children in ages 4 to 6, is caused by pragmatic or semantic reasons. Experimenters would induce children to make the deductive reasoning required by quantifiers like every without using them, and evaluate whether exhaustive pairing would persist.
Bayesian Few Shot Learning of Compositional Instructions (2019)
Developed a Bayesian Model that reproduced human behavior when given the sequence-to-sequence task of interpreting a list of instructions in an artificially generated language to generate a sequence of colors.
Meta-Visualization: Investigating Rapid Learning and Feature Reuse (2019)
Investigated the nature of the meta-learning process in algorithms like MAML through the development of a visualization tool for the learning path in the loss landscape and geometric interpretations of rapid learning and feature reuse.
Implementing a Fusion Tree in C++ (2017)
Implemented a fully functional and well documented fusion tree that can perform predecessor queries with a constant number of operations in a general BigInt.
CodCad (2016)
CodCad was an online platform created to teach competitive programming for free. I co-founded CodCad in 2016.
Noic (2016)
Noic is a project that promotes scientific olympiads in Brazil and democratizes access to them. I presided Noic in 2016
publications
Method and System for an End-to-End Deep Learning Based Optical Coherence Tomography (OCT) Multi Retinal Layer Segmentation Permalink
We use a Transformer-based model to segment retinal layers from OCT scans. We process an image as 1D sequence of A-scans and treat each of them as a token, instead of processing a 2D image, which is more computationally efficient.
Text-image Alignment for Diffusion-based Perception Permalink
We use automatically generated captions to improve the text-image alignment of a diffusion backbone in downstream visual tasks such as semantic segmentation, depth estimation and object detection. Our method also achieves improves the SOTA in both single-domain and cross-domain tasks.
Mice in the Manhattan Maze: Rapid Learning, Flexible Routing and Generalization, With and Without Cortex Permalink
We designed the “Manhattan Maze,” a reconfigurable navigation task that allows systematic map design over a vast space of possible layouts, to study cognitive flexibility in mice. Naive wildtype mice rapidly learned complex maps within two days and accelerated at learning new ones. Acortical mice, lacking the hippocampus and most of the neocortex, learned more slowly at first but still approached optimal performance, suggesting the cortex is not strictly required for navigating the Manhattan Maze.
Humans and Mice Navigate Mazes Alike. Can AI Beat Them? Permalink
We tested human participants on a virtual binary maze game adapted from a prior mouse study and found not only similar performance, but also striking parallels in learning dynamics between humans and mice. We then used this embodied navigation task to compare AI agents with both species, and showed that two canonical agents, a Deep Q-Learning (DQN) model and a Large Language Model (LLM), were outperformed by the biological learners.
Diffusion-Based Action Recognition Generalizes to Untrained Domains Permalink
We propose using features generated by a Vision Diffusion Model (VDM), aggregated via a transformer, to achieve human-like action recognition across domain shifts. We find that generalization is enhanced by the use of a model conditioned on earlier timesteps of the diffusion process to highlight semantic information over pixel level details in the extracted features. Our model sets a new state-of-the-art across three generalization benchmarks, bringing machine action recognition closer to human-like robustness.
Inference-Time Scaling of Diffusion Models via Progressive Seed Pruning Permalink
We propose Progressive Seed Pruning (PSP), an inference-time scaling method for diffusion and flow-matching models that scores intermediate denoised estimates and progressively narrows a large pool of candidate seeds, spending compute on promising trajectories instead of resampling a fixed-size pool. Across diffusion and flow-matching backbones, PSP consistently improves reward-guided selection and achieves higher GenEval scores and better human evaluation on prompt-alignment than best-of-N, importance-sampling, and tree-search baselines at matched compute.
talks
Talk 1 on Relevant Topic in Your Field
Published:
This is a description of your talk, which is a markdown files that can be all markdown-ified like any other post. Yay markdown!
Conference Proceeding talk 3 on Relevant Topic in Your Field
Published:
This is a description of your conference proceedings talk, note the different field in type. You can put anything in this field.
teaching
Informatics Olympiads Teacher
Middle and High School classes, Farias Brito, 2017
I was a teacher of competitive programming for middle and high school students preparing for the Brazilian Olympiad in Informatics and the International Olympiad in Informatics at Organização Educacional Farias Brito.
EE 148b: Large Language and Vision Models
Head Teaching Assistant, Caltech, 2024
Teaching Assistant, Caltech, 2023
I was a Teaching Assistant (2023) and later Head Teaching Assistant (2024) for EE 148b: Large Language and Vision Models at Caltech. This course introduces the foundations, design, training, and evaluation of large language models and vision-language models, with an emphasis on modern multimodal systems, reasoning, alignment, and emerging research directions.
EE 148a: Introduction to Deep Learning
Head Teaching Assistant, Caltech, 2026
I was the Head Teaching Assistant for EE 148a: Introduction to Deep Learning at Caltech. It is an introductory course on the foundations and applications of Deep Learning, with a focus on Computer Vision.