Cornell AI Network Newsletter, October 2026

Cornell AI Network's monthly digest, where we share the latest Cornell AI news, jobs, events, and more.

Join our community: LinkedIn Company Page | cornellainetwork.com | Email us | WhatsApp Group

Launching the "Bridging the AI Gap" program

Cornell AI Network is excited to launch the Bridging the AI Gap program. The mission is to accelerate Cornell University's AI transformation by pairing AI-native students with work experience across departments, professors, and labs, led and mentored by alumni. Learn more about the Bridging the AI Gap here.

If you're an alum and interested in being a mentor or advisor on a project, click here.

If you're a student looking for an apprenticeship or mentorship, click here.

If you're a professor, lab, or department, learn more here.

Announcements

On campus

• Robots for orchards. Cornell is leading a four-year, $7.5 million USDA project to put robots to work in US orchards. Manoj Karkee's team will build robots that pollinate flowers, thin fruit, harvest apples, and weed between rows. The grant also sets up a Center of Excellence for Orchard Robotics in Biological and Environmental Engineering. Gilbert Plath of Washington Fruit and Produce Co. in Yakima says labor now runs over 60% of costs, up from about 45% fifteen years ago.
• Computing with light. Francesco Monticone heads a $7.5 million Defense Department project on the foundations of computing with light. The team will design hybrid optical and electronic systems that could cut the energy and computing needed for AI image analysis. Stanford, Penn, Vanderbilt, and MIT are partners.
• Mathematicians take on AI safety. Two Cornell mathematicians won $3.8 million in grants for AI safety and research integrity. Lionel Levine received $1.5 million from Coefficient Giving to build tools that evaluate the "constitution" of a frontier language model. Daniel Halpern-Leistner received $2.3 million from DARPA to build autoformalization tools, which translate written mathematics into machine-checked proofs.

Upcoming Cornell AI events

• Charvi Rastogi, Google DeepMind. Tuesday, October 27, 2:00 to 3:00 pm, Gates 203 and Zoom. Rastogi's Global AI Seminar talk, "Participatory Localized Red Teaming for Text-to-Image Safety in the Global South," presents PLACES, a set of over 26,000 text-to-image model failures collected with universities in Ghana, Nigeria, and two regions of India.
• Cornell Systems Summit. November 1 to 3, Statler Hotel. Systems Engineering and AI: Frontiers in Research, Education, and Workforce Development brings industry, government, and academic leaders to Cornell for panels and working groups on the AI skills gap in systems engineering. Registration is open.
• AI and Peer Review Panel. Friday, November 13, 12:20 to 1:30 pm, Gates 310 and Zoom. Panelists are Sarah Dean, Paul Ginsparg, Bharath Hariharan, Jon Kleinberg, and Yian Yin. The panel is listed with the Computer Science AI seminars.
• Local AI Hackathon. November 14 at 11:00 am to November 15 at 2:00 pm, Gates Hall room CIS142. Build an app that runs entirely offline, alone or in a team of two. Registration is required.

Featured event: Agostino Capponi, Data Science Distinguished Lecture (November 17)

• Agostino Capponi, professor of industrial engineering and operations research at Columbia, gives the Data Science Distinguished Lecture on Tuesday, November 17, 4:15 to 5:15 pm, in Rhodes Hall 253. A Zoom stream is available.
• His talk, "Data-Driven Learning of Latent Dynamics," presents a method that combines diffusion maps with a geometric Kalman filter. Tested on data from three financial crises, the method cut portfolio return prediction error by up to 52% against standard scenario analysis and up to 57% against dynamic principal component analysis.
• The ORIE Colloquium and the Center for Data Science for Enterprise and Society host the lecture.

Recent events

• Judge John Murphy '99 of the U.S. District Court for the Eastern District of Pennsylvania gave the Distinguished Raymond G. Thorpe Lecture at Olin Hall on October 19. His talk was titled "How a Cornell Chemical Engineer Became a Federal Judge, and the Lessons Learned Along the Way." Murphy holds a Caltech PhD in chemical engineering and serves on the Judicial Conference's Committee on Codes of Conduct and the Subcommittee on Artificial Intelligence.
• The Center for Teaching Innovation, the Einhorn Center, and the Global AI Initiative ran a two-part seminar, Exploring the Ethics of AI in Community-Engaged Learning, on October 6 and 13 in Gates 203. The listing for the second session promised reports from faculty and community partners on how they handle AI ethics questions in practice.

Full events calendar: cornellainetwork.com/events

Company profile from the Cornell network

• ASPEED. Yi-Chen Li '27, a 2026 summer research intern at Cornell's Cañizares Center for Emerging Markets, profiled this Taiwanese chipmaker, founded in 2004. ASPEED makes the baseboard management controllers that let operators run AI servers remotely, and it holds an estimated 70% to 80% of the global BMC market. Revenue more than doubled in 2024 and rose another 40.6% in 2025. Gross margin reached 68% in 2025.

Career opportunities

Featured AI and ML positions:

• Funded PhD positions in AI and science, Vertaix lab, Princeton Computer Science. Fall 2027 entry, full funding, apply through Princeton CS by December 15, 2026. Posted by Adji Bousso Dieng, MPS '13.
• Founding Infrastructure Engineer, Structured AI, New York. Structured AI is a Y Combinator company whose AI agents review construction drawings for architects and engineers. This is the first infrastructure hire on a team of ten engineers. Posted by Ethan Otto, MEng '22.

Full job board: cornellainetwork.com/jobs

AI and ML news by stack layer

• Infrastructure. AI for Medical Imaging (AI4MI), a Weill Cornell Radiology initiative with Cornell Tech, draws on secure AI clusters and Empire AI for its computing.
• Model development. The Artificial Intelligence Materials Institute (AI-MI), the first NSF AI institute for materials science, is hiring postdoctoral fellows to speed up AI-driven discovery of materials for energy, sustainability, and quantum technology. The post pays $80,000 and starts in spring or summer 2027. Applications sent by October 15 got first consideration, and review continues until the posts are filled.
• Application interface. The Cornell AI Platform is a secure, private sandbox in pilot, free, and looking for collaborators. AI Gateway gives access to models from Google, Anthropic, OpenAI, and others, and AI Agent Studio lets users build agents that run on schedules or triggers.
• Orchestration. Haoyang Li, a postdoctoral associate in Weill Cornell's Division of Health Informatics and AI, won a Mastercard and Weill Cornell Medicine seed grant for LabARIA, a multi-agent system for clinical laboratory operations and quality management.
• Cross-layer. Weill Cornell Medicine's FAIMI fellowship funds early-career faculty using AI in biomedical research, clinical care, healthcare delivery, and biomedical education.

STEEPLE: AI through seven lenses

Social, technological, economic, environmental, political, legal, and ethical.

Social

How AI changes behavior, relationships, and access to information.

• Cornell Tech's Emma Harvey argues that AI researchers should treat data annotation as a measurement problem. She notes that 30% of recent annotation papers at leading computational linguistics venues reported no quality management, and she proposes "community face validity assessments" before researchers can use a dataset.
• CATAI, the Cornell Tech initiative on assistive technology and AI, works on assistive robotics and inclusive sensing with partners including ADAPT, YAI, and the Sean Parker Institute for the Voice.
• The Cornell Digital and AI Literacy Initiative designs and tests community-driven tools and toolkits for the technology literacy problems facing young people and vulnerable communities.

Technological

New architectures, tools, and capabilities.

• Linxi Zhao and colleagues in the Cornell NLP group won Best Paper Runner-Up at a NeurIPS 2025 workshop for Limited Memory Language Models. The model stores facts in an external database as well as in its weights, and learns during pre-training to look them up.
• Zizhao Chen and Yoav Artzi built KnotGym, a 3D benchmark for spatial reasoning. Models tied two-crossing knots 83% of the time and three-crossing knots 16% of the time. "Once moved to reason in the 3D world, AI breaks," Chen told the Chronicle.
• Dan Landau, known for an AI-powered liquid biopsy that detects tumor DNA in blood, became inaugural chair of Weill Cornell's Department of Systems and Computational Biomedicine on September 1. "Artificial intelligence applications thrive at larger scales," he said.
• Cornell Tech created a new post for Mert Sabuncu as associate dean for research. Sabuncu works on AI for medical imaging and is vice chair of AI and engineering research in Weill Cornell Radiology.

Economic

Market impacts, business change, jobs, and investment.

• Cornell's Treasury team recovered $100,000 in unidentified payments with AI. The tool matches payments against three years of ledger history and researches vendors through the Cornell AI Gateway. Vendor names appear on 99% of unidentified payments, while invoice and PO numbers appear on fewer than 4%. Departments confirmed five payments from the first batch, and about $1 million in backlog remains.
• At an August workshop, 30 participants from 23 units across eight colleges tested the AWS AI Development Lifecycle method for building software. Many found the hardest problems were not technical. "Building is not what's expensive, and not what's slow," said Ayham Boucher, executive director of the Cornell AI Innovation Hub.

Environmental

AI's carbon cost, climate uses, and sustainability trade-offs.

• The Cornell AI for Sustainability Initiative (CAISI) applies AI to decarbonizing energy and agri-food systems. Its NSF traineeship runs five AI4S courses, among them AI for Energy Systems (CHEME 6830) and AI for Digital Agriculture (CHEME 6840).
• About 80 researchers, including 28 invited speakers from seven countries, met at Cornell Tech in June for a workshop on physics-aware machine learning for molecules and materials. "Ultimately, we want machine learning models that get the right answers for the right reasons," said organizer Shuwen Yue.

Political

Government policy, international competition, and national AI strategy.

• The Cornell Global AI Initiative cites one figure to explain why it exists. Many AI systems are rooted in WEIRD populations (Western, educated, industrialized, rich, and democratic), which are less than 15% of the world's population. Its governance work studies the policies and institutions needed for transparency, accountability, and public oversight.
• Artificial Intelligence, Policy, and Practice (AIPP), led by Solon Barocas, Jon Kleinberg, Karen Levy, and Helen Nissenbaum, studies AI as a social and technical system with computational, legal, mathematical, philosophical, and sociological methods.

Legal

Copyright, liability, privacy, and regulation.

• In a 2023 paper, Gianclaudio Malgieri and Cornell's Frank Pasquale argue that high-risk AI should require a license, with developers justifying a system before deployment rather than after harm.
• Katherine Lee, A. Feder Cooper, and James Grimmelmann trace copyright questions through each stage of the generative-AI supply chain, from training data to deployment. The Cornell Law AI page lists their 2023 paper under featured research.
• Robert A. MacKenzie, one of five new Cornell Law faculty, directs the Entrepreneurship Law Clinic. His work covers the growing role of AI in legal practice and legal education.

Ethical

Bias, fairness, consent, and safety research.

• Six new Cornell Tech faculty arrive between November 2026 and July 2027. Pratyush Maini's benchmarks for machine unlearning, which test whether a model can forget specific data, are now widely used. Ayush Sekhari works on unlearning and privacy.
• Goldwater Scholar Neer Metha '28, a mathematics major, says the award is a chance to "connect applied mathematics with AI to solve real-world problems."

This month's AI signals

What to act on, what to watch, and what to guard against.

Green: developments and tools to try

• Mathematicians can try mathcopilot.org, a free tool built by Cornell's Daniel Halpern-Leistner, whose new $2.3 million DARPA grant funds autoformalization research.

Yellow: trends to watch

• A contested proof. OpenAI said on September 8 that an internal system solved the Navier-Stokes problem. The Clay Mathematics Institute awards a prize only after a proposed solution is published and withstands two years of scrutiny. The Cornell Daily Sun asked four Cornell professors what the claim means for research. "What's surprising is how fast they have gotten good," said Steven Strogatz.
• Management training has not kept up. A CAHRS note on Korn Ferry's "A Management Crisis in the Making" says AI agents have helped make management work about as "foreign and unrecognizable" as it was when the field first emerged in the first half of the 20th century. CAHRS academic director Brad Bell says, "Conventional management training doesn't work."
• Skeptics should take part. Sun columnist Luke Dolan argues that AI skeptics should "fight for a hand on the wheel" instead of refusing the technology, and asks, "Who gets to decide what technological progress means for the people living through it?"

Red: risks and regulation

• ILR's Center for Applied Research on Work launched a two-year pilot, "AI and the New Informality," on whether AI erodes the job protections of professional work in law, finance, journalism, and design. Sarosh Kuruvilla, Virginia Doellgast, and Zoë West will interview professionals in New York, California, Berlin, and Geneva. They plan a diagnostic for how informal a job has become and a governance playbook.

AI term of the month

Agentic systems

An agentic system is an AI system that pursues a goal by planning and taking a sequence of actions across tools, data, and workflows, instead of returning a single answer. Once such a system can act in the real world, it raises questions of oversight, safety, and accountability.

• Stanford's Nigam Shah spoke on "Responsible AI in Healthcare: From RWE to Agentic Systems" in Weill Cornell's Dean's AI Lecture Series on September 1. RWE stands for real-world evidence. The recording is available.
• LabARIA, Haoyang Li's multi-agent system for clinical labs, is one agentic system in development at Weill Cornell.

Related entry: "intelligent agent" in Wikipedia's glossary of artificial intelligence

Research paper of the month

GEMS: Molecular Structure Elucidation as Learned Search over Isomer Graphs

Utku Umur Acikalin, Yingheng Wang, Goncalo J. Gouveia, Tyler J. Schwertfeger, Frank C. Schroeder, Carla P. Gomes | arXiv (cs.AI), October 2026

Identifying an unknown small molecule from its formula and mass spectrum is a bottleneck in metabolomics and natural-product discovery. GEMS starts from candidate structures and refines them with learned bond edits that keep the molecular formula and every atom's valence intact. A neural policy scores all valid edits in one pass. On the NPLIB1 benchmark it ranks the correct structure first 29.6% of the time and identifies a structure in about 3.3 minutes, against 26.2 hours for the authors' FRIGID run. All six authors are at Cornell, in computer science, chemistry, and the Boyce Thompson Institute.

Read the paper

Video of the month

How AI Is Reshaping Global Research and Innovation

Barry Barish and Chuck Ng, introduced by Benjamin Z. Houlton | September 16, 2026

In this eCornell keynote, CALS Dean Benjamin Houlton opens with the college's research priorities. Nobel laureate physicist Barry Barish and entrepreneur Chuck Ng then discuss the roles and challenges of AI in global science. eCornell asks for a name and email before the recording plays.

Watch on eCornell

Community opportunities

Get involved

Weekly AI office hours. Every Thursday at 10 am ET, open to Cornell AI project teams. Ask for the link at contact@cornellainetwork.com.

Open volunteer roles. Newsletter section editors and event coordinators. Introduce yourself on LinkedIn.

Cornell AI resources

• Cornell AI Initiative
• Cornell AI Tools & Resources
• Wikipedia's Glossary of Artificial Intelligence
• eCornell AI and Tech Keynotes
• Cornell AI Network Job Board

Featured alumni spotlight

Two of the co-first authors of BINN, a biogeochemistry-informed neural network for soil carbon are former Cornell researchers. Joshua Fan, a former doctoral student in Carla Gomes's computer science lab, now works on Chrome AI agents at Google. Feng Tao, a former Schmidt AI for Science postdoctoral fellow at Cornell, is now an assistant professor at Nanyang Technological University in Singapore.

• BINN couples a soil carbon model with a neural network and learns from 25,925 soil profiles across the contiguous United States.
• It runs more than 50 times faster than the earlier PRODA method, with similar accuracy and less spatial bias.
• Soil holds about three-quarters of the world's land-based carbon.
• Yiqi Luo's lab in the School of Integrative Plant Science led the work, with Gomes and CALS Dean Benjamin Houlton as co-authors.

About Cornell AI Network

Cornell AI Network is Cornell graduates and affiliates who work on AI. Write if you have a project, a job, or a correction.

Connect with us: contact@cornellainetwork.com | LinkedIn | cornellainetwork.com | WhatsApp Group

Want to contribute content, suggest topics, or share opportunities? Reply to this newsletter or reach out to contact@cornellainetwork.com.