AIYGO Newsletter
Edition One - August 2025 to August 2026
Introduction
Hello everyone. This is the AI Youth Global Organization’s (AIYGO) first newsletter!
It has been approximately one year since the founding of the AIYGO, and we are incredibly proud of where it has ended up today. A big thank you to everyone who has made the organization what it is today and for your time and effort.
As I’m sure you all know, the AIYGO is a largely collaborative organization and relies on the abilities of everyone - not just one individual - and so everyone has played a crucial role in AIYGO’s development as an organization.
In this newsletter, we will be covering projects, upcoming events, member accomplishments, and how you can get involved.
We hope you enjoy the newsletter!
Best,
Francesco H., US ITC Researcher, and Felipe Q., US Senior Chair
I 2025-2026 Event Recap
Early Days (August, 2025 to October, 2025)
The AI Youth Global Organization (AIYGO) was founded by students from the United States, Mexico, South Korea, New Zealand, and Nepal who crossed paths at the awards ceremony for the International Mathematical Modeling Challenge (IMMC) in Hong Kong.
Now, if you know anything about mathematical modeling, you will know that math modeling is simply applied mathematics and computer science. Similarly, artificial intelligence is simply applied mathematics and computer science. Putting the two together, we thought, why not apply our mathematical modeling schools to artificial intelligence education and research.
And so, on October 22nd, 2025, the AI Youth Global Organization was founded. Their goal was simple. To give high schoolers not only a voice but an influence in the field of Artificial Intelligence, a field that will likely become the most important field of the 21st century.
And this goal was built on two pillars. One, to make rigorous AI education accessible to high schoolers around the world, and, two, to create an international research network serious enough to make a positive impact in the field of machine learning.
Educational Resource Development and Growth (December 2025, to March, 2026)
From October to December, the organization leaders were quite busy with school, but nonetheless attempted to meet every two weeks, develop online educational resources — notably the Neural Networks from First Principles resource, and draft the very first version of our organization charter. At the time, our network remained very small and limited to the mathematical modeling competition that we had all met each other at.
In December, however, as we neared the Winter break and a few of our Senior Chairs discovered LinkedIn, the organization saw a period of unprecedented growth. From December to late January, the organization developed four new educational resources, covering topics including Physics Informed Machine Learning and Transformers, developed partnerships with organizations such as Non-Trivial, secured three new guest speakers, and saw the admission of about 10-15 new Senior Chairs.
From February to March, the organization saw continued growth as about 10-15 new Senior Chairs and ITC Researchers joined from around the world, and developed four new educational resources covering topics including Physics Informed Neural Networks, World Models (e.g., JEPA), Attention Variants (e.g., GQA, MLA, and KDA), and Reinforcement Learning with Human Feedback. Our new charter was voted in as well, formalizing many organization operations including the admission of new members, the ⅔ majority voting process for matters involving the core organization, and the overall future success of the organization. In early March, realizing that we had a talented group of students in the organization who were all eager to learn, we began to form research teams to apply our newly acquired knowledge.
Continued Growth and Summer Build Up (May, 2026 to early June, 2026)
From April to May, the organization continued to see steady growth, with about 5-10 new Senior Chairs and ITC Researchers joining the organization. Following AP Exams in early May, however, the organization once again saw a period of significant growth. In May, about 20 new Senior Chairs and ITC Researchers joined the core organization, and the core organization began to coordinate an online startup school and research school. Unfortunately, we were only able to find two (fantastic) speakers — Jonathan Tremblay, a Senior Staff Research Scientist at NVIDIA and Abhishek Divekar, a Senior Applied Scientist at Amazon — for our research school, which unofficially took place in early June. However, our startup school saw significantly more success as you will see in the section.
It was also at this time that the organization began to finalize the directions of its research projects for the rest of the summer.
AIYGO Startup School and Learning (July, 2026)
In early July, AIYGO hosted its first online startup school, where we hosted speakers such as Cory Levy — the Founder of Z Fellows, Divya Narendra — the founder of SumZero, five Y Combinator Founders, two Z Fellows, one a16z speedrun founder, and an Investor at Entrepreneurs First. The full details can be seen on the event info website, and the video recordings will be released soon — unless requested otherwise by the speaker.
Additionally, during July, the organization began its intense learning period to fully develop the research projects they began in June. Some of the topics learned included Polynomial Chaos Expansions, Cellular Sheaves from “Towards a Spectral Theory for Cellular Sheaves” by Jakob Hansen and Robert Ghrist, Flow Matching and Diffusion Models, a little bit of Compiler Theory (specifically the Chomsky Hierarchy), and some advanced Reinforcement Learning frameworks.
Undergraduate Division, the International Mathematical Modeling Challenge, and Research (August, 2026)
In early August, two of the AIYGO’s Co-Founders had the honor of presenting at the International Mathematical Modeling Challenge awards ceremony hosted at Harvard University. The video presentation for the event can be seen here.
Throughout August the organization members were also working extensively on their projects listed in Section II. Note that not all of the projects within the organization have been listed in Section II.
Finally, in August, the AIYGO began to set up its core undergraduate division, where its members will be establishing university circles at Cambridge University, MIT, Columbia, Harvard, UPenn, the University of Oxford, and Stanford.
II Projects
Note that not all of the projects within the organization have been listed in Section II.
Language Models for Low-Resource Natural Language Processing
Our language modeling committee is currently working on distilling “linguistically informed” weak priors that probabilistically model language structure into transformers via soft constraints. The purpose of such a construction is to help the transformer overcome its tendency to overfit in low-resource regimes and bias its optimization towards linguistically plausible regions of the parameter space. Our preliminary framework used a Hidden Markov Model, which is mathematically equivalent to probabilistic regular grammars, as a weak structural prior over outputs, showing a 3-5% improvement in the perplexity that needs to be verified further across Nahuatl, Quechua, and Guaraní. Our current work centers on introducing additional weak structural priors, notably Probabilistic Context-Free Grammars (PCFGs), which are more expressive than probabilistic regular grammars within the Chomsky hierarchy and are capable of capturing hierarchical structures in language.
PhysRNA-Filter
AlphaFold 3 and RoseTTAFold All-Atom predict protein-RNA binding poses but no representation of the thermodynamic cost of adopting that pose from free solution. We introduce a Physics-Informed Graph Transformer (PhysGT). PhysGT is a Siamese graph transformer for RNA-protein mutation ∆∆G prediction that goes beyond 4-term linear contact energy by learning from graph structure and language model embeddings.
Cellular Sheaves as Validators for Generative Material Models
This team is currently investigating cellular sheaves because of their gluing properties that enforce global consistency. Because of this global consistency condition, we are investigating the use of cellular sheaves as validators for generative materials models (and possibly other generative machine learning models) that often struggle with consistency at a global scale — specifically, the Dirichlet energy, which can be derived from the sheaf Laplacian and measures the degree to which local signals violate compatibility relations induced by the sheaf's restriction maps. We also believe that there may be a way to parameterize the restriction maps characteristic of cellular sheaves with a neural network u, where we can optimize the parameters of that neural network to meet certain physical or geometric constraints.
Hail Risk for Solar Energy Portfolios
This team is currently working in collaboration with Munich Re, the number one reinsurance firm in the world, to model hail risk for solar energy portfolios. Currently, the team is taking a hybrid approach, combining constraint-based optimization for region generation, Bayesian statistics for conditional likelihoods of hail events, extreme value theory for expected hail size distributions, and advanced kinetics to model the expected damage on solar panels.
Financial Agents
For the past few years, developing AI agents that are capable of making informed trading decisions has been of great interest to hedge funds, quantitative trading firms, and other financial agents. However, creating a quantitatively informed agent is a task of great difficulty as traditional language model agents struggle with properly analyzing large amounts of data to determine what is and what is not a signal. Additionally, approximating the necessary gradient for a “good” versus “bad” financial action is not something that can be easily defined by a differentiable loss function as the AI agent. Therefore, this team is taking a hybrid approach by unifying the outputs of various foundation models and indicator data into one unified signal space that an AI agent can act on. The team is using the Reinforcement Learning with Verifiable Rewards approach to optimize the parameters of the agent, and is currently researching potential reward functions.
III Member Accomplishments (Fall 2025 to Summer 2026)
Note, the following list of accomplishments is just a list of accomplishments of our members. These accomplishments did not necessarily have anything to do with the organization, but we would like to acknowledge them for their incredible achievements.
Olympiads
Math Olympiad
Anango P., Member of the US AIYGO-UD: International Math Olympiad, Bronze Medal
Daniel L., Canada ITC Researcher on Trial: USA Mathematical Olympiad, Qualifier
Physics Olympiad
Eric Z., US ITC Researcher: United States of America Physics Olympiad (USAPhO), Silver Medal
Chemistry Olympiad
Ishan M., Member of UK AIYGO-UD: Royal Society of Chemistry United Kingdom Chemistry Olympiad (RSC UKChO), Gold Medal
Eric Z., US ITC Researcher: United States National Chemistry Olympiad (USNCO), High Honors (T50 Nationally)
Ryan M., US ITC Researcher on Trial: United States National Chemistry Olympiad (USNCO), Study Camper (T20 Nationally)
Artificial Intelligence Olympiad
Anango P., Member of the US AIYGO-UD: International AI Olympiad (IOAI), Gold Medal + Rank 6
Roumak D., India Senior Chair: International AI Olympiad (IOAI), Silver Medal
Savyasachi S., India Senior Chair: International AI Olympiad (IOAI), Silver Medal
Aayush Y., Nepal Senior Chair: International AI Olympiad (IOAI), Honorable Mention
Anango P., Member of the US AIYGO-UD: International Winter AI Olympiad (IAIO), Absolute Winner
Aaditya B., US ITC Researcher: International Winter AI Olympiad (IAIO), Honorable Mention
Himaneesh S., India ITC Researcher on Trial: India International AI Olympiad Training Camp (IOAITC), Attendee + Top 10
Raj S., India ITC Researcher on Trial: India International AI Olympiad Training Camp (IOAITC), Attendee + Rank 1 in ML Theory
Informatics Olympiad
Anango P., Member of the US AIYGO-UD: International Informatics Olympiad (IOI), Silver Medal
Roumak D., India Senior Chair: India International Informatics Olympiad Training Camp (IOITC), Attendee
Roumak D., India Senior Chair: Indian National Olympiad in Informatics, Gold Medal
Trisanu D., India ITC Researcher on Trial: Indian National Olympiad in Informatics, Bronze Medal
Dennis Y., US ITC Researcher on Trial: United States of America Computational Olympiad, Finalist
Roumak D., India Senior Chair: United States of America Computational Olympiad, Platinum Division
Linguistics Olympiad
Shinkyeom B., UAE Senior Chair: International Linguistics Olympiad (IOL), Team Leader
Economics Olympiad
Eli O., US ITC Researcher on Trial: International Economics Olympiad (IEO), Gold Medal + Rank 4
Liqian Y., China ITC Researcher on Trial: International Economics Olympiad (IEO), Gold Medal + Rank 9
Hyuk W., South Korea ITC Researcher on Trial: International Economics Olympiad (IEO), Gold Medal + Rank 16
Other Competitions
Mathematical Modeling
Ananya A., Singapore ITC Researcher on Trial: International Mathematical Modeling Challenge (IMMC), International Round Meritorious
Anubhav P., Nepal Senior Chair: International Mathematical Modeling Challenge (IMMC), International Round Meritorious
Catherine X., Singapore ITC Researcher on Trial: International Mathematical Modeling Challenge (IMMC), International Round Meritorious
Daniel L., Canada ITC Researcher on Trial: International Mathematical Modeling Challenge (IMMC), International Round Outstanding
Himangsu A., Nepal Senior Chair: International Mathematical Modeling Challenge (IMMC), International Round Meritorious
Hui N., Singapore ITC Researcher on Trial: International Mathematical Modeling Challenge (IMMC), International Round Meritorious
Yanxi S., Singapore ITC Researcher on Trial: International Mathematical Modeling Challenge (IMMC), International Round Meritorious
Abhay M., Member of AIYGO-UD, Former US ITC Researcher: Mathworks Math Modeling Challenge, Finalist
Physics Tournaments
Jaehyun W., South Korea Senior Chair: International Young Physicists' Tournament (IYPT), Gold Medal + International Rank 2
Aarush D., US ITC Researcher: United States of America Young Physicists Tournament (USAYPT), 1st Place
George Z., US Senior Chair: United States of America Young Physicists Tournament (USAYPT), 1st Place
Michael G., US Senior Chair: United States of America Young Physicists Tournament (USAYPT), 1st Place
Chemistry Tournaments
Ishan M., Member of UK AIYGO-UD: International Chemistry Tournament (IChTo), Silver Medal
Science Fairs
Leonardo P., Brazil Senior Chair (Temporarily on Leave): International Science and Engineering Fair (ISEF), 4th Place Grand Award
Leonardo P., Brazil Senior Chair (Temporarily on Leave): International Science and Engineering Fair (ISEF), 4th Place Special Award
Liqian Y., China ITC Researcher on Trial: International Science and Engineering Fair (ISEF), Finalist
Warren F., Member of US AIYGO-UD: International Science and Engineering Fair (ISEF), 2nd Place Grand Award
Warren F., Member of AIYGO-UD: Regeneron Science Talent Search (STS), Scholar
Zeviel P., US ITC Researcher: Regeneron Science Talent Search (STS), Scholar
Investment Competitions
Eralp K., Turkey Senior Chair: Wharton Global High School Investment Competition, Semifinalist
George Z., US Senior Chair: Wharton Global High School Investment Competition, Semifinalist
Economics Competitions
Eli O., US ITC Researcher on Trial: National High School Fed Challenge, Winner
Business Competitions
Sai B., Member of AIYGO-UD, US ITC Researcher: DECA International Career Development Conference, 2nd Place — Hospitality and Tourism Professional Selling
Programs
Aaditya B., US ITC Researcher: MIT PRIMES
Aaditya B., US ITC Researcher: Ross Mathematics Program
Adithya R., UK ITC Researcher: European Summer Program on Rationality (ESPR)
Eric Z., US ITC Researcher: Garcia Summer Research Program
Himaneesh S., India ITC Researcher: Emergence AI, Intern
Leonardo P., Brazil Senior Chair (Temporarily on Leave): Research Science Institute (RSI)
Logan D., Singapore ITC Researcher on Trial; Research Science Institute (RSI); Top 5 Presenter
Michael G., US Senior Chair: NASA, Intern
Roumak D., India Senior Chair: Emergence AI, Intern
Universities
Ishan M., Member of UK AIYGO-UD: Cambridge University
Oliver P., UK ITC Researcher, Member of UK AIYGO-UD: Cambridge University
Vihaan S., Member of US AIYGO-UD: Carnegie Mellon University
Francesco H., Member of US AIYGO-UD: Harvard University
Anango P., Member of US AIYGO-UD: Massachusetts Institute of Technology
Manu I., US ITC Researcher: Massachusetts Institute of Technology
Zeviel P., US ITC Researcher: Massachusetts Institute of Technology
Anay A., US ITC Researcher on Trial: Stanford University
Jayveer K., US ITC Researcher on Trial: Stanford University
Abhay M., Member of AIYGO-UD, Former US ITC Researcher: University of California, Berkeley, Management, Engineering, and Technology
Shubh J., Member of US AIYGO-UD: University of California, Berkeley, Management, Engineering, and Technology
Michael L., New Zealand Senior Chair: University of New South Wales
Adithya R., UK ITC Researcher: University of Oxford, Ellison Scholar
Alston Y., New Zealand Senior Chair: University of Oxford, Ellison Scholar
Giorgio S., Member of UK AIYGO-UD: University of Oxford, Ellison Scholar
Maxmilian E., Member of UK AIYGO-UD: University of Oxford, Ellison Scholar
TianLe L., Advisor: University of Pennsylvania, School of Engineering
Sai B., Member of AIYGO-UD, US ITC Researcher: University of Pennsylvania, Wharton
Frank H., Member of US AIYGO-UD: Yale University
IV Upcoming Events
On September 1st, Petar Veličković, a Senior Staff Research Scientist at Google DeepMind will be speaking to the organization members from 12 pm to 1 pm Eastern Standard Time.
By early September, most of the teams will have completed the base version of their projects that they may choose to continue building over the next year.
An updated version of the Charter will be sent out soon to vote on.
Our undergraduate division will begin the formation of university circles at various universities.
We will have likely some new Senior Chairs joining us from Canada and Singapore in the upcoming weeks.
V Additional Announcements
Logic Islands: Logic for the AI Age
You’re on an island inhabited by two kinds of people: truth-tellers, who always tell the truth, and liars, who always lie. You need directions; but first, you need to determine whom you can trust.
You meet three residents, Ari, Bolo, and Cala:
Ari: Bolo is a truth-teller.
Bolo: There are more truth-tellers than liars among the three of us.
Cala: At least one of the three of us is a liar.
Just from their statements alone (which may be true or false), can you determine which of them are telling the truth?
AIYGO recently hosted mathematician and game designer Daniel Rosiak, author of Sheaf Theory through Examples, for a presentation on Logic Islands: Deduction Game, a mobile game built around hundreds of handcrafted logic challenges like this one, with ambitions to become “the app for logic in all its glory.”
As you go from island to island, the puzzles begin with approachable truth-teller-and-liar problems (like above) before developing into mind-bending challenges involving incomplete information, strategic questioning, nested beliefs, and reasoning about what others can know. In some challenges, you must determine exactly which question to ask and to whom to ask it.
Without ever feeling like a textbook, Logic Islands also doubles as a hands-on introduction to logic itself. It essentially begins with propositional reasoning and gradually leads players through predicate logic and beyond, into modal and epistemic questions about what different agents know, believe, and can infer.
In an age when AI systems can produce fluent, confident answers almost instantly, it matters more than ever to be able to distinguish what is genuinely supported by the available information from what merely sounds plausible. In such an age, knowing logic will be a superpower. Logic Islands makes the discipline of logic and challenging and intuitive as possible.
Daniel is looking forward to interacting more with the AIYGO community, and we hope to hold an organization-wide Logic Islands Challenge in early 2027. Players will compete to reason through a handpicked set of island puzzles as efficiently as possible. So download the game and come ready to compete!
Play Logic Islands: Deduction Game on iOS or Android: