About Danqi,

I am currently a Ph.D. student in the Computer Science department at Yale University, advised by Prof. Smita Krishnaswamy.


My research interests include geometric, manifold learning, and deep learning. Specifically, I am interested in studying the characteristics that make expressive and generalizable embeddings and developing efficient and principled methods to achieve such representations. I am also interested in AI/ML for bioscience.


I graduated Magna Cum Laude from Northwestern University and obtained my master's degree from Princeton University. While at Princeton, I worked on computer perception and video recognition in Prof. Olga Russakovsky's lab.


Before Princeton, I worked as a software engineer at Meta for two years.

Publications

Geometry-Aware Generative Autoencodersfor Warped Riemannian Metric Learning and Generative Modeling on Data Manifolds

Xingzhi Sun *, Danqi Liao *, Kincaid MacDonald, Yanlei Zhang, Guillaume Huguet, Guy Wolf, Ian Adelstein,Tim G. J. Rudner, Smita Krishnaswamy

AISTATS, 2025 |ICML GRaM, 2024 | arXiv

DiffKillR: Killing and Recreating Diffeomorphisms for Cell Annotation in Dense Microscopy Images

Chen Liu *, Danqi Liao *, Alejandro Parada-Mayorga *, Alejandro Ribeiro, Marcello DiStasio, Smita Krishnaswamy

ICASSP, 2025 | arXiv, 2024

Assessing Neural Network Representations During Training Using Noise-resilient Diffusion Spectral Entropy

Danqi Liao*, Chen Liu*, Ben Christensen, Alexander Tong, Guillaume Huguet, Guy Wolf, Maximilian Nickel, Ian Adelstein, Smita Krishnaswamy

CISS, 2024 | ICML TAG-ML, 2023

Sentence Embeddings using Supervised Contrastive Learning

Danqi Liao

Arxiv Preprint, 2021

Selective Feature Aggregation for Single Frame Supervised Temporal Action Localization

Danqi Liao

Master Thesis, 2022

Projects

Attention mechanism on GPU

Danqi Liao

My CUDA c implementation of forwad pass of regular Attention and FlashAttention.

Peer to Peer Proxy Network

Danqi Liao

P2P proxy network aims to protect user privacy and hide their online identity by forwarding their network requests randomly.

Kademlia Distributed Hash Table

Danqi Liao

My implementation for Kademlia DHT, a distributed hash table for decentralized peer-to-peer computer networks.

Teaching (as TA/Head TA)

  • CPSC452: Deep Learning Theory and Applications , Yale, Spring 25
  • CPSC474: Computer Vision and Biological Perception, Yale, Fall 24
  • COS429: Computer Vision, Princeton University, Fall 21
  • COS326: Functional Programming, Princeton University, Fall 20
  • EECS324 Operating System, Northwestern University, Spring 18
  • EECS321 Programing Language, Northwestern University, Fall 17
  • COS324: Advanced Programming Technique, Princeton University, Spring 22
  • COS226: Data Structures & Algorithms, Princeton University, Spring 20