Interdisciplinary by Design: What High School Student AI Research Really Looks Like

Since 2022, our students have published or presented research at more than 100 professional conferences and journals, applying AI advances to domains from medicine to music. The range isn't an accident. Here is what interdisciplinary high school research looks like.

Which Forums and What Student Grades?

• More than 20 different highly respected PhD-reviewed conferences and workshops, including IEEE, SPIE, ACM, AAAI, SAE, NeurIPS Climate, ICML, CVPR, MIT URTC, and others. These students have presented alongside professors, PhD researchers, and industry professionals.

• Students from 7th to 12th grade. Many students who started early had multiple publications prior to high school graduation.

Problems first, methods second

Interdisciplinary research means more than pointing an algorithm at a dataset. To ask a question worth answering, mentors guide students to reach into their interests, passions, and coursework to identify a problem that they care about. For example, a paper on glaucoma screening has to start from how the disease shows up in a retinal scan, while a paper on mosquito surveillance has to start from how mosquitoes sound and where they live, and a paper on violin practice has to start from how violinists learn. Each of these is drawn from the student's own interests and experiences, guided by our mentors. The result is a record that spans a long list of domains. As examples, our student publications have covered the following:

Health and medicine: glaucoma, cataract, and diabetic retinopathy screening, retinopathy of prematurity in newborns, cervical, colorectal, and breast cancer detection and recurrence prediction, glioblastoma and stomach cancer survival modeling, brain MRI analysis, Alzheimer's, epilepsy, leukemia diagnosis, early sepsis diagnosis, neonatal exposure to maternal smoking, pharmacogenetics, fibromyalgia, antibiotic resistance, parasite egg detection, and mental health in children.

Drug discovery: compounds and inhibitors for conditions including hypertension, cancer, and cardiac fibrosis, drug-target interaction, allergy prediction with protein models, and toxin-binder screening for antivenom discovery.

Agriculture: pest detection in corn, rice, and beets, weed detection in soy, apple disease classification, shelf life estimation for produce, pollinator classification for cocoa, and irrigation scheduling.

Environment and sustainability: forest fire surveillance and wildfire spread prediction, flood mapping, underwater plastic detection, waste sorting, parking and carbon reduction, textile sustainability, power plant emissions from remote sensing data, mycoremediation, mosquito surveillance from images, sound, and edge devices, and solar panel tilt optimization.

Space: solar flare detection, galaxy redshift prediction, star and constellation recognition.

Transportation and safety: pothole detection from dashcam video, early driver-risk prediction, vehicle battery state-of-charge prediction, underwater acoustic classification of vessel propulsion sources, and network intrusion detection.

Accessibility and assistive technology: wearables for visual assistance and tools for people with visual impairments

Music, arts, education, and society: violin performance learning, Kulintang music transcription, online dance education, attributing artist influence in AI art, teaching strategies for children with autism, vaccine sentiment analysis, anxiety prediction from online chat, and volleyball action detection.

That breadth comes from students following their own questions instead of being steered toward one narrow topic. It also reflects something true about AI today: nearly every field is being reshaped by it, so the skill that matters is knowing how to apply it to a real problem.

Depth shows up in the record too. Mosquito surveillance, for example, appears in four papers over about two years, moving from smartphone images, to sound, to low-power audio detection, to a system built for edge devices. The field kept moving, and so did the work.

Judged by Experts

Many of the venues are specialist meetings in health, optics and imaging, music audio, climate, automotive engineering, and edge computing. This means that papers are typically reviewed by independent experts in those areas, people the student has never met and who have no reason to go easy.

That matters most for interdisciplinary work. It's one thing to say a paper bridges AI and medicine, and another to have it accepted at a venue where specialists in that field are reading. A grade comes from a teacher who knows the student, and a competition score comes from a judge in the room. A conference acceptance comes from reviewers who read the work cold and decide it met the standard. It is the same review process professional researchers go through, which is why the people presenting before and after the student are often professionals and PhDs.

Other researchers are also building on the work. One student's paper on making the textile industry more sustainable, for example, has already been cited 10 times on Google Scholar. These citations, by people who do not know the student, are further evidence of direct impact.

State Of The Art (SOTA) AI Methods

The AI behind these papers is current and varied. The record includes transformer-based models, large language models, ChemBERTa and ProteinBERT for chemistry and biology, protein language models, ensembles of convolutional and transformer architectures, generative AI, 3-D convolutional networks, survival analysis, transfer learning, reinforcement learning and physical AI, low-power weakly supervised learning for audio, machine learning on small embedded microcontrollers, and the YOLOv8 and YOLOv10 object detectors. No single technique runs through the whole collection. Our mentors guide students to pick the method that fits the question.

This is why the work matters beyond the paper. The students are learning serious AI skills, hands-on and in context: preparing real data, choosing and training models, evaluating results honestly, and writing it up for expert readers. Those are the skills that serve students well in internships, in college coursework and research labs, and in whatever they pursue after that.

Beyond the Publication

Getting a paper accepted is one achievement. Presenting it is another, and it teaches things a paper can't. We teach students how to present their work in a professional venue, how to network at the event, and how to connect with other researchers. Mentor support doesn't end when the paper is submitted. Mentors guide students on their presentation decks, rehearse the talk with them, and run mock Q&A as needed, so the hardest questions aren't a surprise on the day.

A student who has stood in front of strangers in their field, answered their questions, and talked with other researchers afterward has done something most students their age haven't.

Some of our students have told us it was a valuable learning experience, and some have written about it in their college essays. It also shows up in interviews. A student who has had to explain the work to experts understands it solidly, and that helps them explain the research clearly in college and internship interviews, where the questions tend to go past the results and into the why.

We know this model first-hand. Every AIClub founder holds a PhD, and we know that graduate students heading into a talk get practice sessions, deck reviews, and repeated feedback from faculty until they're ready. We give our students the same kind of support.

Book a free consultation here.