NYU Center for Data Science

10/23/2026

CDS is hiring Faculty Fellows for fall 2027! 📣

Fellows will lead research projects of their choosing and work with faculty across CDS and NYU.

We welcome postdoctoral researchers working in core machine learning/AI methods, machine learning theory, AI for science, computational social and political sciences, and responsible AI.

The two-year, full-time appointments offer a competitive annual base salary, benefits, and funds for research and travel, with a teaching load of one CDS course per year.

Applicants must hold a doctorate by the planned start date of September 1, 2027, and should have a strong publication record and experience teaching quantitative topics.

Apply by November 24, 2026, for full consideration: https://apply.interfolio.com/193498

10/23/2026

🧪 A useful chemistry result can be buried deep inside a paper.

That’s why AskChem was created: to make individual findings searchable, linking each one to its source evidence.

The AskChem team included:

🔬 NYU Computer Science PhD candidate Bing Yan
🔬 CDS Professor of Computer Science and Data Science Kyunghyun Cho
🔬 CDS-affiliated Associate Professor of Physics, Chemistry, Mathematics and Neuroscience Stefano Martiniani
🔬 Matterstack coauthor Gregory Wolfe

Researchers can compare reaction conditions, find measurements, and follow links between supporting or conflicting findings.

In a 30-question test, GPT-5.5 using AskChem cited paper identifiers that all resolved to real records, compared with 88.3% without a retrieval tool.

Read more: https://nyudatascience.medium.com/askchem-a-new-search-engine-for-chemistry-research-50d28404b692

10/23/2026

ATTENTION, Prospective Master's Students! A reminder that the master's program admissions information session is taking place THIS MONDAY, October 26, 10am ET on Zoom.

Take a step forward in building your future career in data science. Register now: https://nyu.zoom.us/webinar/register/WN_oIhnlp-RTT66MtZtPJL0wQ #/registration

10/22/2026

Large language models store information as long lists of numbers, not as words or symbols, yet they handle math, logic, and code. A new preprint offers an account of how.

Led by Yale linguist Tom McCoy with CDS Associate Professor of Linguistics and Data Science Tal Linzen, Paul Soulos, and Paul Smolensky, the study found that these vectors closely approximate symbolic structure. When the team swapped models’ internal representations for symbolic ones that pair “fillers” (words) with “roles” (subject, object), behavior stayed largely the same across seven LLMs working in language, arithmetic, logic, and code.

The researchers could also edit those representations directly. Given “the clever doctor helped a lawyer,” they moved “clever” to the object, and the model behaved as if it had read “the doctor helped a clever lawyer.”

Paper: https://arxiv.org/abs/2608.29530

Yale News: https://news.yale.edu/2026/10/08/how-do-ai-chatbots-reason-humans-yale-led-study-offers-clues

The neural networks that power popular AI chatbots process data much differently than humans but capably perform tasks involving language, logic, and math. A new study sheds light on how those systems “think.”

10/21/2026

A new AI test offers a faster, cheaper way to estimate whether breast cancer will return.

CDS Visiting Scholar Krzysztof J. Geras and collaborators developed Ataraxis Breast, which uses routine pathology slides and basic clinical data to estimate recurrence risk within a day rather than weeks.

In a study of 8,161 patients across seven countries, Ataraxis Breast successfully separated patients at higher and lower risk of recurrence and did better than the widely used genomic test Oncotype DX in a direct comparison of 858 patients.

The 51-author Nature Communications paper also included several researchers with ties to CDS:

🔬 CDS Associate Professor Carlos Fernandez-Granda
🔬 CDS founding director Yann LeCun
🔬 CDS PhD alum Kangning Liu
🔬 CDS MS alum Ken G. Zeng

The research began at NYU and became the basis for Ataraxis Breast, the first clinical test developed by Ataraxis AI.

https://nyudatascience.medium.com/ai-test-offers-a-faster-cheaper-way-to-predict-breast-cancer-recurrence-4988c1a82bd2

Some of the most widely used tests for estimating whether breast cancer will return were developed some 25 years ago and require a lab to…

10/20/2026

The CDS Alumni Council recently hosted a finance and fintech industry skills panel on careers for data scientists and machine learning professionals.

The panel covered risk management, algorithmic decision-making, fraud detection, and transaction analytics, alongside advice on networking, internships, and career growth.

Thank you to our alumni panelists:

🎤 Quantitative Researcher Sabrina Zhu, Kepos Capital
🎤 Partner Sarvesh Patki, Konik Capital Partners
🎤 Data Analyst Ademola Oladosu, SRS Investment Management LLC
🎤 Principal Data Scientist Tatenda Ndambakuwa, Morgan Stanley

The discussion was moderated by Data Scientist Sarath Kareti, CNN.

10/19/2026

CDS Assistant Professor of Computer Science, Engineering, and Data Science Emily Black told MS Admissions Ambassador Mehul Goel that the mindset she appreciates most in her classes is "a bit of self-direction."

She said she throws a lot at students and hopes they find a thread to pull on for their project, though a full research project isn't required to do well.

"My hope is that people are coming, sort of excited to do something on their own," she said.

"I think it's a class where you get out what you put in."

10/16/2026

World models can plan by imagining what will happen next — but those plans can break when the real environment differs from training.

CDS PhD student Ying Wang and her collaborators, including CDS founding director Yann LeCun and CDS Asst. Prof. Mengye Ren, developed AdaJEPA, which lets a JEPA world model learn from what actually happens after each action and update itself before planning again.

Across tests involving unseen object shapes, visual changes, altered dynamics, and new maze layouts, the approach improves planning with lightweight test-time updates.

In one low-data experiment, adaptation increased success from 28.1% to 60.8%, surpassing a frozen model trained with 16 times as many trajectories per shape.



https://nyudatascience.medium.com/a-world-model-that-keeps-learning-after-training-ends-77b366643499

10/16/2026

CDS recently welcomed Domino Data Lab for an info session on internships and new grad roles in forward deployed engineering and software engineering.

The session brought students together with Domino’s team, including two recent NYU graduates.

Thank you to our speakers:

🎤 Forward Deployed Engineer Bella Chang (NYU ’26)
🎤 Forward Deployed Engineer Praveen Bandla (NYU ’26)
🎤 Campus Recruiter Bianca Tulloch

Learn more about Domino’s student and new grad opportunities: https://domino.ai/careers/campus-recruiting

10/15/2026

Attention, Prospective Master's students! You won't want to miss our upcoming admissions MS program information session on Monday, October 26, 10:00am.

The session will feature talks from Julia Kempe, Director of Graduate Studies for MSDS Program, Isabella Paris, Graduate Program Administrator, and Zoë Levin, Assistant Director of Career Development & Industry Relations.

Register now: https://nyu.zoom.us/webinar/register/WN_oIhnlp-RTT66MtZtPJL0wQ #/registration