Math meets medicine
We believe our research can help bring affordable world-class medical expertise to every patient around the world
Under the Microscope
Start using the most advanced AI infrastructure built solely for healthcare.
Core research
Model reasoning
General purpose LLMs are powerful at solving math exams but fail when they are tasked with complicated real-world healthcare reasoning.
Learning to reason from a complex healthcare history, the most recent medical research, and laboratory results takes a specialized reasoning agent with embedded knowledge of millions of related healthcare cases.
Policy optimization
Policies define the protocol followed by a machine learning model. Policies are learned from human guidance and general purpose LLM policies are defined by non-healthcare-trained individuals.
Optimizing policies through feedback from healthcare professionals enable learning the protocol from the best in class.
Alignment & interpretability
AI systems are inherently bad at knowing when they don’t know. Building interpretability into the AI allows healthcare professionals to understand the reasoning process and locate the source of its output.
Alignment AI models are tasked with interpreting the output of other AI models at scale, flagging inconsistencies before they harm.
Latest publications
Computer vision, Adaptive scale, Vision Transformers
2024-10-10
MSViT: Dynamic Mixed-Scale Tokenization for Vision Transformers
Explainability
2024-08-15
Normalized AOPC: Fixing Misleading Faithfulness Metrics for Feature Attribution Explainability
Explainability, Automated medical coding
2024-06-13
An Unsupervised Approach to Achieve Supervised-Level Explainability in Healthcare Records
Retrospective study, Stroke, Classification, ASR
2023-12-19
A Retrospective Study on Machine Learning-Assisted Stroke Recognition for Medical Helpline Calls
Automated medical coding, Review
2023-07-23
Automated medical coding on MIMIC-III and MIMIC-IV: A Critical Review and Replicability Study
Self-supervised learning, Speech Recognintion, Automatic...
2022-03-21
Self-Supervised Speech Representation Learning: A Review
Out-of-distribution detection, Uncertainty quantification
2022-03-02
Model-agnostic out-of-distribution detection using combined statistical tests
Self-supervised learning, Speech representations
2022-03-01
A Brief Overview of Unsupervised Neural Speech Representation Learning
Workshop paper
2022-02-22
Benchmarking Generative Latent Variable Models for Speech
Automatic speech recognition Self-supervised learning...
2021-06-06
On scaling contrastive representations for low-resource speech recognition
Automatic speech recognition
2021-02-17
Do We Still Need Automatic Speech Recognition for Spoken Language Understanding?
Automatic speech recognition, Contextualized representations...
2021-02-17
Do end-to-end speech recognition models care about context?
Variational auto encoder, Variatonal inference...
2021-02-16
Hierarchical VAEs Know What They Don't Know
Question segmentation Automatic speech recognition...
2020-05-12
MultiQT: Multimodal learning for real-time question tracking in speech
Workshop paper
2019-10-16
Towards Hierarchical Discrete Variational Autoencoders
Conference paper
2019-02-06
BIVA: A very deep hierarchy of latent variables for generative modeling
Workshop paper
2018-11-28
On the Inductive Bias of Word-Character-Level Multi-Task Learning for Speech Recognition
Workshop paper
2017-12-01
Utilizing Domain Knowledge in End-to-End Audio Processing
Workshop paper
2017-04-03
Semi-supervised generation with cluster-aware generative models
Variational autoencoder, Variatonal inference
2016-02-17
Auxiliary deep generative models
Preprint
2016-02-16
How to train deep variational autoencoders and probabilistic ladder networks
Variational autoencoder Variatonal inference...
2016-02-16
Ladder variational autoencoders
Preprint
2015-09-17
Recurrent spatial transformer networks
Want to do research with us?
We are always looking for brilliant minds dedicated to innovating in healthcare, as well as impact research collaborations. Get in touch if you want to work together.
Join Our Mission
We believe everyone should have access to medical expertise, no matter where they are.