CURRENT WORK
My current work mainly centers on causal structure learning and causal reasoning. My research is motivated by urgent biomedical problems, including drug discovery and fairness in healthcare. As a Research Intern in Machine Intelligence at Microsoft Research Cambridge, I explored causal reasoning evaluation and elicitation in LLMs.
My research interests include:
- Probabilistic and causal graphical modeling.
- AI reasoning evaluation and elicitation.
- The challenges that arise in low-data settings.
- The automation and acceleration of scientific discovery and clinical research.
A subset of my recent works are highlighted below. The rest of my publications can be accessed through Google Scholar.
|
|
Compositional Causal Reasoning.
Under review. 2024. J Maasch, A Huyuk, X Xu, A Nori, J Gonzalez.
[arXiv coming soon]
[tl;dr] We explore the ability of AI agents to infer how causal quantities compose and propagate through causal graphs.
|
|
Local Causal Discovery for Structural Evidence of Direct Discrimination.
Under review. 2024. J Maasch, K Gan, V Chen, A Orfanoudaki, N Akpinar, F Wang.
[
arXiv
]
[
code
]
[
slides
]
[tl;dr] Local structure learning facilitates causal fairness analysis in complex decision systems, as illustrated by a real-world case study on organ transplant allocation.
A preliminary version of this work was presented as an invited talk at INFORMS 2024.
|
|
Local Discovery by Partitioning: Polynomial-Time Causal Discovery Around Exposure-Outcome Pairs.
Uncertainty in Artificial Intelligence. 2024. J Maasch, W Pan, S Gupta, V Kuleshov, K Gan, F Wang.
[
arXiv
]
[
code
]
[
poster
]
[
slides
]
[tl;dr] Local causal discovery enables efficient automated covariate selection under limited prior knowledge with guarantees on correctness.
Preliminary versions were presented as a poster at the NeurIPS 2023 Causal Representation Learning Workshop and as an invited talk at POMS 2024.
|
|
Molecular de-extinction of ancient antimicrobial peptides enabled by machine learning. Cell Host & Microbe 31. 2023. J Maasch*, M Torres*, M Melo, C de la Fuente. *Equal contribution.
[
paper
]
[
code
]
[
npr
]
[
nature news
]
[
cnn
]
[
vox
]
[tl;dr] Machine learning guides bioinspired prospection for encrypted antimicrobial peptides that display in vitro and in vivo activity with low host toxicity.
|
PROFESSIONAL ACTIVITIES
INVITED TALKS
10.24 | INFORMS Annual Meeting | Seattle, WA | Local Causal Discovery for Structural Evidence of Direct Discrimination |
07.24 | Microsoft Research Machine Intelligence Core | Cambridge, UK |
06.24 | University of Cambridge Statistical Laboratory | Cambridge, UK | Local causal discovery for effect estimation |
04.24 | 34th Annual POMS Conference | Minneapolis, MN | Local Discovery by Partitioning |
ACTIVITIES In 2023, I co-organized Cornell's interdepartmental Causal Reading Group, a student-led discussion group on causality and causal inference. I co-developed CS 6006: Succeeding in the Graduate Environment, the first-year PhD seminar at Cornell Computer Science. With students at Cornell, MIT, and NJIT, I co-organized the 2024 NYC meetup for the Learning on Graphs Conference, Learning Meets Geometry, Graphs, and Networks. I am a student leader for the PhD admissions and recruitment process at Cornell Tech.
PEER REVIEW [Computing] International Conference on Artificial Intelligence and Statistics (AISTATS); Association for Computational Linguistics (ACL) Rolling Review; ICML Workshop on Structured Probabilistic Inference & Generative Modeling (SPIGM); NeurIPS Women in Machine Learning Workshop (WiML). [Life sciences] Communications Biology (Nature Portfolio); Bioinformatics (Oxford Academic); ACS Infectious Diseases (American Chemical Society).
PRIOR WORK
As a master's student at Penn Engineering, I used machine learning to investigate the therapeutic potential of antimicrobial peptides. My master's research was supported by the University of Pennsylvania's GAPSA-Provost Fellowship for Interdisciplinary Innovation and the Reproducible Research Fellowship, funded by the Open Knowledge Foundation and Alfred P. Sloan Foundation.
Prior to this, I was a research assistant in various clinical and experimental biology labs. As an experimentalist, I helped develop molecular diagnostics for neglected tropical diseases in academia and for viral pathogens in industry.
CONTACT
Preferred contact: In general, I can be reached through LinkedIn.
Pronouns: they / them.
If you are new to non-binary pronouns, here are some examples for how to use them in a grammatical way from Merriam Webster, the MLA Style Guide, and the APA Style Guide. ☺
echo @ | sed 's/^/maasch/' | sed 's/$/cs.cornell.edu/'
|
Updated October 2024. This website was adapted by J. Maasch from this source code.
|
|