PROJECT GALLERY
Student AI research projects
Every project here was scoped, built, and published or presented by a Veritas AI student. Filter by track to see work in your area of interest.
CLIMATE
Forecasting Harmful Algal Blooms in Tampa Bay
Harmful algal blooms like Florida's Red Tide kill thousands of fish and endanger public health. Neel built an XGBoost model that forecasts chlorophyll-a — the key indicator of algal biomass — days in advance from water, weather, and prior bloom conditions, plus a Random Forest classifier that flags whether a bloom is coming.
Published in the National High School Journal of Science
Neel · AI Fellowship, Summer 2025 · Read the paper
HEALTHCARE
Detecting Lung Cancer in CT Scans with a CNN
Slow detection timelines let cancer develop before results come back. Srihari built a convolutional neural network that identifies non-small-cell lung cancer in patient CT scans, cutting the time it takes to flag a case for review.
Published in the American Journal of Student Research
Srihari · AI Fellowship, 2025 · Read the paper
FINANCE
Stock Prediction — and Honestly Measuring Whether It Works
Online claims of highly accurate AI stock prediction usually go untested. Quan forecast prices for the S&P 500, NVIDIA, and JPMorgan across five horizons, then built a rigorous evaluation framework — walk-forward backtesting, a one-year hindcast, and live paper trading — showing that most reported accuracy reflects market drift, not real forecasting skill.
Open-source on GitHub
Quan · AI Fellowship, 2026 · View on GitHub
NLP
A Writing Assistant That Builds Skills Instead of Replacing Them
Most AI writing tools fix grammar but flatten creativity. Lan fine-tuned an LLM with custom data pipelines to build an assistant that sparks ideas through character-driven narratives and scene illustrations — and can adapt to a specific author's style, from Steinbeck to children's books.
Published in the National High School Journal of Science
Lan · AI Fellowship, Summer 2024 · now studying Data Science at UC San Diego · Read the paper
DEEP LEARNING
Auditing Bias in ML Predictions of School Performance
Do predictive models work equally well for every school? Using standardized test data from Ontario's EQAO, Daniel built regression models to predict school performance and ran subgroup error analysis — finding that lower-performing schools consistently receive less accurate predictions, even with no demographic variables in the model.
Published in the American Journal of Student Research
Daniel · AI Fellowship, Winter 2025 · [[BLOCKER — needs the public AJoSR URL; the link on file is a private Google Doc]]
HEALTHCARE
An Interpretable Framework for single nucleotide polymorphism-based Amyotrophic Lateral Sclerosis risk prediction through supervised Machine Learning
HEALTHCARE
Detecting Lung Cancer in CT Scans with a CNN
Slow detection timelines let cancer develop before results come back. Srihari built a convolutional neural network that identifies non-small-cell lung cancer in patient CT scans, cutting the time it takes to flag a case for review.
Published in the American Journal of Student Research
Srihari · AI Fellowship, 2025 · Read the paper
HEALTHCARE
An Interpretable Framework for single nucleotide polymorphism-based Amyotrophic Lateral Sclerosis risk prediction through supervised Machine Learning
STUDENT OUTCOMES
Where our students go next
Veritas AI students have gone on to study at:
PROGRAM BROCHURE
See what your project could be
The brochure covers every program, the mentorship model, and how students get from a first line of Python to work like the projects above.
Program-by-program breakdown, timelines and eligibility
Sample student research projects
How need-based financial aid works, up to 100%
Download our brochure
Every program, what students build, and how financial aid works. No cost, no obligation.
We'll email the brochure and occasional program updates. Unsubscribe anytime.
Your project could be next
Applications are reviewed on a rolling basis. Need-based financial aid is assessed inside the same form.