Agentic AI Program Projects

The same problems top companies solve, built by high schoolers.

Our AI students don't follow tutorials. They pick a real industry challenge, the kind of thing Uber, Amazon, or JP Morgan actually spends engineering resources on, and build a working multi-agent AI system that solves it. Four weeks. Fully open-sourced on GitHub.

w

Projects tackle the same challenges faced by

GoogleAmazonMicrosoftUberJP Morgan ChaseESPN
9
Real-World Projects
9
Industries Covered
4 Wks
To Build & Deploy
100%
Open Source
Full Project Catalog

Every project. Every line of code.

Nine projects across nine industries. Each one tackles a challenge that real companies invest serious engineering resources in, and each one is fully open-sourced on GitHub so you can read every line, fork it, and make it yours.

Intermediate
Advanced
Expert
AI + HumanitiesExpert

Adaptive Language Tutoring System

Inspired byDuolingo

Diagnoses learner weaknesses across six skill dimensions: grammar, vocabulary, reading, listening, writing, and speaking. Uses Item Response Theory to adapt difficulty in real time, builds personalized curricula that spend 60%+ of time on weak areas, and maps everything to the CEFR proficiency framework (A1 through C2).

Item Response Theory6 skill dimensionsCEFR mappingSpaced repetition
AI + Social GoodExpert

Climate Weather Prediction & Disaster Alert System

Inspired byGoogle DeepMind

Three-agent pipeline that forecasts extreme weather up to 10 days ahead. Ingests real atmospheric data (ERA5/GFS), runs autoregressive inference to detect cyclones, heat waves, and cold snaps on a 0 to 100 severity scale, then dispatches emergency alerts in CAP XML format, the same standard used by FEMA and the Red Cross.

GraphCast GNNAutoregressive inferenceCAP XML alertsERA5/GFS ingestion
AI + E-CommerceAdvanced

E-Commerce Product Recommendation System

Inspired byAmazon

Merges explicit preferences with behavioral signals like browsing history, purchases, searches, and cart activity to deliver personalized recommendations. One agent builds a customer affinity profile, another scores the catalog through a multi-signal model, and a third picks a diverse final set while generating plain-language explanations for every suggestion.

Multi-signal scoringBehavioral fusionExplainable outputDiversity ranking
AI + Software DevAdvanced

GitHub PR Review Agent

Inspired byGitHub

Four agents that automate code review on pull requests. Parses diffs, classifies files by language, then runs two agents in parallel: one scanning for security issues (hardcoded secrets, SQL injection, XSS) with CWE references, the other checking style rules. A final agent deduplicates findings, ranks by severity, and posts inline comments.

Parallel agentsCWE referencesMulti-languageInline commenting
AI + HealthcareExpert

Healthcare Prior Authorization System

Inspired byUnitedHealthcare

Reviews clinical data for prior authorization decisions on diagnostic imaging (MRI/CT). Extracts symptoms, diagnoses, and treatment history from unstructured notes, scores each request against ACR Appropriateness Criteria on a 0 to 100 scale, and generates professional determination letters with APPROVED, DENIED, or PENDED decisions.

ACR Criteria scoringClinical NLPUnstructured note parsingDetermination letters
AI + MarketingIntermediate

B2B Lead Scoring & Email Sequence System

Inspired byHubSpot

Scores inbound B2B leads against an Ideal Customer Profile and generates personalized three-email nurture sequences. Enriches each lead with firmographic and behavioral signals, classifies them as Hot, Warm, or Cold, then tailors messaging by industry pain points, company size, and buying intent.

ICP lead scoringFirmographic enrichmentEmail sequencesIntent classification
AI + SportsIntermediate

Soccer Sports Analytics System

Inspired byESPN

Generates post-match summaries and pre-match win probability predictions. Ingests raw match data, produces natural-language reports covering key events and tactical observations, and computes outcome probabilities using team form, head-to-head history, standings, and squad availability.

Win probabilityMatch summariesTactical analysisBroadcast-ready output
AI + TransportationAdvanced

Surge Pricing Agent System

Inspired byUber

Four agents simulate real-time ride-hailing supply and demand across five city zones. Computes surge pricing multipliers from demand/supply ratios, recommends driver reallocation between zones, and produces plain-English reports with full reasoning, making every pricing decision transparent and auditable.

Supply/demand simSurge computationDriver reallocationAuditable reports
AI + FinanceAdvanced

Fraud Detection Co-Pilot

Inspired byJP Morgan Chase

Three agents that flag, investigate, and summarize suspicious transactions. A Detector screens with rule-based and ML-driven signals. An Investigator does a deep dive with contextual evidence. A Summarizer compiles structured case reports, with a human checkpoint for high-severity cases before any final call.

Rule + ML hybridHuman-in-the-loopEvidence gatheringCase reports

Want to build something like this?

Apply Now
What You Walk Away With

A project admissions officers remember.

Your kid won't just “learn AI.” They'll walk out with a working system they designed, a GitHub repo they own, a program certificate, and a story that actually stands out on a college application.

9
Shipped Projects
Deployed & open-sourced
9
Industry Domains
Finance to Healthcare
4
Weeks to Ship
From day one to deployed
1:3
Mentor Ratio
Industry engineers, not TAs
Apply for the Next Cohort

Limited spots. Rolling admissions.