LaunchLoop Wins National AI Challenge 2025 with Revolutionary Exam Grading System
Team LaunchLoop, featuring Echofold CEO Kevin Collins, claims first place at Ireland's National AI Challenge with GradGenie—a production-ready AI system that saves €48 million annually
What is the National AI Challenge?
The National AI Challenge is Ireland's premier AI hackathon competition, organised by TechIreland as part of the National AI Meet. The 2025 edition ran for two weeks, culminating in final presentations on 18 September 2025, where teams competed to develop innovative AI solutions for real-world problems. Team LaunchLoop won first place with GradGenie, a production-ready AI exam grading system that achieves 90% cost savings (€48M annually) for Irish taxpayers.
TL;DR
- •Team LaunchLoop wins National AI Challenge - First place victory at Ireland's premier AI hackathon on 18 September 2025
- •€48 million annual savings - GradGenie reduces Leaving Cert correction costs from €53M to €5M (90% reduction)
- •1 week vs 26 days - Revolutionary 5-agent AI system with 96-97% OCR accuracy processes 61,000 exams faster
- •Production system beats prototypes - Complete AWS infrastructure with security, RBAC, and audit logs whilst competitors delivered MVPs
- •LaunchLoop community strength - Experienced founders from Echofold's LaunchLoop demonstrate collaborative innovation model
On 18 September 2025, Team LaunchLoop claimed first place at Ireland's National AI Challenge, the country's premier artificial intelligence hackathon competition. The winning entry? GradGenie—an AI-powered exam grading system that could transform how Ireland assesses 61,000 Leaving Certificate students annually, saving Irish taxpayers €48 million per year.
Whilst most hackathon teams delivered prototypes and minimum viable products, Team LaunchLoop built something radically different: a complete, production-ready system with enterprise-grade AWS infrastructure, a 5-agent AI grading architecture, multiple redundant OCR models, comprehensive security features, and role-based access portals. Led by Kevin Collins, CEO of Echofold, the team demonstrated how AI automation can solve national-scale challenges.
This victory showcases the power of LaunchLoop, Echofold's community of startup founders who collaborate whilst running their own ventures. In this comprehensive article, we'll explore how Team LaunchLoop won the National AI Challenge, the technical architecture behind GradGenie, and what this means for the future of AI-powered assessment in Ireland.
01.How GradGenie Actually Works: The Technical Reality
GradGenie isn't a simple "AI grades exams" solution. It's a sophisticated three-stage pipeline combining optical character recognition, multi-agent AI systems, and human oversight—all running on production AWS infrastructure designed for national-scale deployment.
Stage 1: Industrial-Scale Scanning
The system begins with bulk scanning using just 13 industrial scanners operating at 120 pages per minute. Running 16 hours daily for 7 days, these scanners provide capacity for 10,483,200 pages against the Leaving Cert's 9,808,800 page workload—maintaining a 7% buffer with 96,343 pages of daily headroom.
Stage 2: Transformer-Based OCR
Once scanned, GradGenie uses multiple OCR models for redundancy and accuracy. The primary system employs transformer-based, GPU-accelerated OCR that achieves 96-97% accuracy over handwritten text. The system transcribes pages in just 0.5 seconds with automatic queuing and performance management.
Stage 3: 5-Agent AI Grading Workflow
This is where GradGenie truly differentiates itself. Rather than a single AI model making grading decisions, the system employs five specialised agents:
Grader Agent (GPT-5)
Evaluates student answer against marking scheme, assigning initial scores with detailed reasoning
Reviewer Agent (Claude)
Checks if grading is fair and unbiased, identifying potential scoring inconsistencies
Decider Agent (GPT-5)
Makes final decision on grade based on grader assessment and reviewer feedback
Validator Agent (Rules)
Approves auto-grading or flags for human review based on 85% confidence threshold
Output Decision
Either auto-grades with full explainability or flags for Later Human Review (LHR)
This human-in-the-loop approach ensures quality whilst maintaining speed. The system processes exams at 2.5 seconds per paper, but any assessment below 85% confidence automatically routes to a human examiner.
02.The €53 Million Problem They Solved
Every year, Ireland's Leaving Certificate examination creates a massive logistical and financial challenge for the State Examinations Commission. Currently, 4,000 teachers spend 26 days correcting exams at up to €10,000 per examiner, costing taxpayers over €53 million annually.
The Challenge By The Numbers
Students taking exams annually
Annual correction costs
Days for full-time correction
The challenge extends beyond cost. Since the pandemic, Ireland has experienced continuous delays in releasing exam results, largely attributed to examiner shortages. These delays create stress for students, parents, and schools awaiting critical information for CAO college applications.
Recognising this challenge, the State Examinations Commission offered a €100,000 contract in April 2025 to researchers studying the use of generative AI in educational assessment. Team LaunchLoop saw this research contract and decided to go further—rather than just studying the problem, they'd build the solution.
03.90% Cost Savings: €48 Million For Irish Taxpayers
GradGenie represents approximately 90% cost savings compared to traditional exam correction. Here's the financial comparison:
| System | Annual Cost | Timeline | Resources |
|---|---|---|---|
| Traditional | €53 million | 26 days | 4,000 teachers |
| GradGenie | €5 million | 7 days | 13 scanners + AI |
| Savings | €48 million (90%) | 73% faster | 99% less labour |
These savings could fund additional teachers, school resources, technology infrastructure, or student support programmes across Ireland's education system. The multi-agent system doesn't just save money—it improves grading quality through consistent marking standards, bias detection, and detailed explanations for every grade.
04.The LaunchLoop Team: Collaborative Innovation
Team LaunchLoop consists of experienced startup founders from Ireland's New Frontiers programme, each running their own ventures whilst collaborating on innovative projects:
Kevin Collins (Team Lead)
CEO of Echofold, specialising in AI automation and agentic workflows for businesses. This was Kevin's first hackathon participation, though he had previously run a workshop on "How to Win a Hackathon."
Nikita Akella
Experienced entrepreneur focused on technology innovation
Alwin Stephen
Founder of NeuraModal, deep technical expertise
Daniel Smyth
Founder of Tendr, product and business development
Vishranth CV
Startup founder with enterprise software background
"We are a collective of experienced founders, each running our own ventures, who came together to solve this challenge."
This diversity of expertise—AI engineering, cloud architecture, product development, business strategy—enabled Team LaunchLoop to build a comprehensive solution far beyond typical hackathon scope. Learn more about joining the LaunchLoop community.
05.Production-Ready System: What Made The Difference
Most hackathon entries deliver proof-of-concept demos or minimum viable products. Team LaunchLoop built something fundamentally different: a production-ready system with enterprise-grade infrastructure. This distinction won them the National AI Challenge.
❌ Typical Hackathon Entry
- • Prototype or MVP demo
- • Single model solution
- • Local development setup
- • Basic security considerations
- • Limited scalability
✓ GradGenie Production System
- • Complete production deployment
- • 5-agent architecture + multiple OCR models
- • AWS auto-scaling infrastructure
- • WAF, encryption, RBAC, audit logs
- • National-scale capacity
Enterprise Security Features
- Web Application Firewall (WAF) - Protects against SQL injection, XSS attacks
- AES-256 Database Encryption - Exam data unreadable if exfiltrated
- Role-Based Access Control (RBAC) - 7 distinct user roles with granular permissions
- CI/CD with SAST - Static Application Security Testing integrated into deployment
The judges understood that GradGenie wasn't a concept—it was a deployable solution. This distinction transformed the team from hackathon participants to serious contenders for government partnership in solving Ireland's exam correction challenge.
06.Frequently Asked Questions
What is the National AI Challenge Ireland?
The National AI Challenge is Ireland's premier AI hackathon competition, organised by TechIreland as part of the National AI Meet. The 2025 edition challenged teams to develop innovative AI solutions over a two-week period, culminating in final presentations on September 18th, 2025. The hackathon brings together Ireland's top AI talent, startup founders, and technology innovators to tackle real-world problems using artificial intelligence.
How does GradGenie work to grade exams with AI?
GradGenie uses a sophisticated 5-agent AI architecture to grade handwritten exams. First, transformer-based OCR models transcribe handwritten answers at 96-97% accuracy in 0.5 seconds per page. Then, five specialised AI agents work in sequence: a Grader Agent (GPT-5) evaluates the answer, a Reviewer Agent (Claude) checks for fairness and bias, a Decider Agent finalises the grade, a Validator Agent ensures quality (flagging anything below 85% confidence for human review), and finally outputs either an auto-grade or flags for human oversight. The system can process exams at 2.5 seconds per paper with full explainability.
Who are the members of Team LaunchLoop?
Team LaunchLoop consists of experienced startup founders from Ireland's New Frontiers programme. The team is led by Kevin Collins (CEO of Echofold), with members including Nikita Akella, Alwin Stephen (NeuraModal), Daniel Smyth (Tendr), and Vishranth Chinnivakkam Vijayakumar. LaunchLoop itself is a community of startup founders dedicated to helping each other succeed, with members running their own ventures whilst collaborating on innovative projects.
How much money does GradGenie save compared to traditional Leaving Cert grading?
GradGenie delivers approximately 90% cost savings compared to traditional exam correction. The current Leaving Cert system costs €53 million annually, requiring 4,000 teachers working 26 days at up to €10,000 per examiner. GradGenie reduces this to approximately €5 million total (€2.5M infrastructure, €1.2M scanning, €0.8M AI operations, €0.5M support), saving Irish taxpayers roughly €48 million per year whilst reducing correction time from 26 days to just 1 week.
What made LaunchLoop's hackathon entry different from other teams?
Team LaunchLoop won because they built a complete production-ready system whilst other teams built prototypes and MVPs. In just two weeks, they created full AWS production architecture with auto-scaling infrastructure, multiple OCR models including transformer-based OCR, a 5-agent grading system, comprehensive security features (WAF, database encryption, RBAC), role-based access portals for students, examiners, parents, and administrators, audit logs, CI/CD pipelines with SAST, and file integrity monitoring. The judges recognised this wasn't a hackathon demo—it was a deployable national-scale solution.
The Future of AI-Powered Assessment Is Here
Team LaunchLoop's victory at the National AI Challenge 2025 demonstrates what's possible when experienced founders collaborate on real-world problems. GradGenie isn't just a hackathon project—it's a production-ready solution that could save Irish taxpayers €48 million annually whilst improving the speed, consistency, and transparency of Leaving Certificate exam correction.
The system combines cutting-edge AI automation technology with practical deployment considerations—security, RBAC, human oversight, and audit trails. This balance between innovation and pragmatism won the National AI Challenge and positions GradGenie as a serious contender for government partnership.
Beyond the immediate application to Leaving Cert grading, GradGenie's modular architecture can extend to school homework correction, Junior Cert exams, and international markets. The agentic workflows are language-agnostic, making adaptation to UK A-Levels and European markets straightforward.
Transform Your Business with AI Automation
Echofold specialises in AI automation and agentic workflows for businesses ready to leverage AI at scale. Whether you're exploring AI adoption or scaling existing implementations, our team brings the same production-focused approach that won the National AI Challenge.