When Students Build the Future: GDG on Campus × OpenTPI at Computex 2026 InnoVEX

Nine open-source projects. Four universities. One unmistakable signal: the next generation of AI-native developers is already here.

There is a specific kind of electricity in the air at Computex InnoVEX — a stage reserved for innovation, for the bold, and for the truly new. This year, that electricity will carry a distinctly younger charge. Between industry veterans and funded startups, student developers from four Taiwanese universities (NCU, NCCU, NCUE, NTPU) are set to take the InnoVEX stage. They will showcase what they had quietly been building: a collection of open-source AI applications, each one grounded in real human pain points, each one powered by the digiRunner Open Source Project (OpenTPI) as its API gateway backbone.

These weren’t demo toys or academic exercises. They were working systems — systems that reveal a generation of builders who think natively in agents, gateways, LLMs, and RAG pipelines. As the plan lead behind the OpenTPI project, I couldn’t be more proud, more moved, and frankly more inspired by what these students have created and shared, freely, with the world.

Let’s walk through each of them!

Figure: Overview of Open Source AI Projects by GDG on Campus

🏦 QuantDashboard AI — NCU

Turning market noise into plain-language insight for retail investors

Built by students at National Central University (NCU), QuantDashboard AI tackles one of the most relatable frustrations in personal finance: the ordinary investor who stares at KD, MACD, and candlestick charts without knowing what any of it actually means. The system routes all API calls through a digiRunner gateway, pulls real-time prices and historical K-line data, and feeds the raw numbers into a Dify AI workflow that generates human-readable analysis — no jargon, no guesswork. Automated price alerts and email notifications mean you no longer need to be glued to the screen. For engineers, the Docker-containerized, gateway-governed architecture is a textbook example of how to build scalable, observable financial data pipelines with open-source tooling.

🌲 Magic Pinecone — NCU

A one-stop AI campus portal for the fragmented university experience

Every university student knows the feeling: announcements in one system, course registration in another, events somewhere else entirely, and none of it works well on mobile. Magic Pinecone reimagines the campus information experience as a unified Flutter frontend, with digiRunner handling centralized routing and security, a FastAPI backend doing background syncing from NCU’s various data sources, and a RAG engine (vector database + LLM) providing personalized, accurate answers to natural-language questions. Smart course advising, personalized bulletin summaries, and a genuinely modern UI — all open-source, all composable. This is what a next-generation campus portal should look like.

📚 Course Selection Assistant — NCU

AI that actually helps you pick the right class — and prevents schedule conflicts

Course registration is a semester ritual that shouldn’t require a spreadsheet, three browser tabs, and a lot of frustration. The Course Selection Assistant integrates with NCU’s Portal SSO for identity verification, routes everything through digiRunner (which actively masks student IDs and blocks malicious traffic), and then lets students ask Dify AI questions in plain language — “recommend a light elective that doesn’t conflict with Thursday afternoons.” The system cross-references syllabi, ratings, and schedules automatically. For the school, it means secure academic data integration and dramatically reduced load on administrative staff handling routine course queries.

🧭 Campus Navigation API — NCU

Voice-driven, conversationally intelligent wayfinding for a campus that has its own language

Traditional maps don’t know that NCU students call certain buildings by informal shorthand. This project does. Users speak a destination aloud (“I need to get to Mgmt 2”), and the system processes the voice input through a Web Speech API, routes through digiRunner’s security shield, uses Gemini AI to extract navigational intent and resolve campus-specific aliases from a local SQLite database, calls Google Maps for optimal routing, and synthesizes the result back into natural, humanized spoken directions. Hands-free, localization-aware, and modular enough to be extended to any campus anywhere. The architecture diagram alone is worth studying.

🏥 AI Medical Communication Training Platform — NCU

A gamified, voice-interactive simulator for high-pressure clinical conversations

This one is genuinely moving. Nursing staff face some of the most emotionally demanding communication challenges in any profession, yet training opportunities for difficult patient and family conversations are scarce and often low-fidelity. This NCU project creates a risk-free clinical sandbox: Gemini AI generates randomized, dynamic patient scenarios with different emotional profiles and family demands; ElevenLabs neural voice synthesis makes the virtual patients sound real; and digiRunner keeps the whole multi-AI pipeline stable and secure. Trainees speak, the system responds, and mistakes become learning moments rather than real-world harm. Gamified. Immersive. Potentially transformative for healthcare workforce development.

🎓 AI-Native Scholarship Hub — NCUE

From two hours of paperwork to 20 seconds of AI — a 99.8% efficiency leap

National Changhua University of Education (NCUE) students built something that has already attracted attention from the Ministry of Education and gained traction at NCKU and Feng Chia University. Their Scholarship Hub replaces the traditional process of drafting announcements, manually checking eligibility, and sending notifications with an AI-native workflow: digiRunner handles secure routing, Dify’s Agentic RAG engine parses scholarship conditions, Supabase performs automated student matching, and multi-channel push notifications reach eligible students proactively. A process that took two hours now takes 20 seconds. The project is 100% open-source on GitHub under PolyForm Noncommercial 1.0.0 and is already multi-campus ready.

💳 AI Credit Card Optimization System — NCCU

Real-time checkout recommendations that fill a gap no existing product has closed

Students at National Chengchi University (NCCU) identified something specific and valuable: while tools like iCard.AI, Honey, and ShopBack exist, none of them deliver a genuine real-time AI decision at the exact moment of checkout — they either do static lookups or issue coupons. This system crawls card benefit data, uses LLM-powered rule extraction to structure the rewards logic, deploys a strict calculation engine to eliminate hallucinations, and integrates a Chrome extension for seamless checkout-context detection. digiRunner orchestrates the background workflows and job scheduling with financial-grade reliability. The result: instant, personalized card recommendations with zero cognitive overhead for the shopper.

⚖️ PatentMind AI — NCCU

Secure, private RAG for patent attorneys navigating office action responses

Patent law sits at the intersection of extreme technical complexity and extreme time pressure. NCCU’s PatentMind AI addresses the three compounding challenges patent attorneys face when responding to office actions: manual synthesis across dozens of documents, the logical rigor required to argue non-obviousness, and statutory deadlines where a single missed date means lost rights. The system uses digiRunner’s gateway for data masking (ensuring sensitive case files never leave the firm’s control), Dify AI for RAG-based logical analysis against on-premise patent databases, and automated deadline alerting. For enterprise IT, the Docker-containerized private RAG environment is a compliance-ready, auditable deployment model that other LegalTech applications should study closely.

🎓 Reinventing Tutoring with AI — NTPU

An AI-powered SaaS that turns a cram school into a data-driven growth engine

Taipei National University of the Arts? No — this one comes from National Taipei University (NTPU), and it addresses a problem that is enormous in Taiwan’s education market: the cram school industry’s twin crises of teacher burnout and invisible student churn. The system provides 24/7 automated grading, Socratic-style guided questioning (to prevent students from simply outsourcing thinking to AI), an analytics dashboard with churn detection and early-warning alerts, and intellectual property protection for proprietary course materials. For the tutoring center, the efficiency gain is 80% reduction in grading time. For the operators, deep LMS integration creates the kind of switching costs that generate stable subscription revenue and, built on open-source infrastructure, scales elegantly.

A Message to the Global Developer Community

What ties all nine of these projects together is not just the digiRunner API gateway — though that shared infrastructure is a genuine testament to the OpenTPI project’s vision of democratizing enterprise-grade API governance for student developers. What ties them together is a design philosophy: identify a real human pain point, architect a layered solution, build in security and governance from day one, and share the work openly with the world.

These students — from NCU, NCCU, NCUE, and NTPU — have demonstrated that the Agentic AI era doesn’t belong exclusively to well-funded labs or large engineering teams. It belongs to the builders. It belongs to anyone with the curiosity to ask “what if,” the skills to build it, and the generosity to open-source the result.

The revolution in agentic AI development thinking isn’t just coming. It’s already running on a FastAPI server, routed through digiRunner, somewhere on a university campus in Taiwan — and now, it’s coming to COMPUTEX.

If you are a developer, an enterprise IT architect, an educator, or simply someone who believes that open collaboration is how the best ideas compound, we want to hear from you.

Catch Us Live at COMPUTEX 2026! Come meet these young builders, see their projects in action, and explore how the digiRunner open-source community is shaping the future. Come build with us.

• Dates: June 2 — June 5, 2026
• Location: InnoVEX Zone, Open Source Team Taiwan Pavilion
• Exclusive Developer Task: Stop by our booth to chat about your “Agentic AI nightmares,” star our OpenTPI project on GitHub, and unlock exclusive open-source stickers and limited-edition swag!

Stay updated with the latest OpenTPI news and events: https://www.linkedin.com/company/opentpi