Letting students build AI bots risked generating harmful content. Now, real-time safety models filter all outputs for 330k+ learners.
A nationwide educational program teaches AI literacy to over 330,000 students aged 12 to 16 through an immersive simulation where they configure bots to generate social media posts and influence a fictional election.
Allowing young students to freely prompt generative models posed severe risks of producing hate speech or age-inappropriate content, which would...
“Enabling us to create immersive, engaging learning experiences is a game changer in preparing young people for the reality of the world they live in. We are making sure young people are learning critical thinking and ethical reasoning in ways that they choose to engage with.”
Free AI education program and STEM resources for Australian K-10 students.
Cloud computing services, AI infrastructure, and data analytics platforms for enterprises.
Developed the Win the Farm simulation and built the secure infrastructure for the program.
Related implementations across industries and use cases
Server costs held back scale. A serverless AI platform now delivers adaptive tutoring to 10,000 students, cutting idle spending to zero.
Classrooms with 1:30 ratios left students behind. Now, 350M monthly queries route to specialized models for personalized tutoring.
Class sizes left the "middle 80%" of students unseen. AI agents now coach learners, surfacing gaps for immediate human intervention.
High-latency APIs slowed the platform. Moving to Gemini and GKE cut deployment from days to hours, accelerating global expansion.
Generic models failed safety checks for students. Now, a compliant assistant explains math with building blocks instead of numbers.
Filming and editing took months. Teams now bypass the camera, using AI avatars to generate and localize content in hours.
Students navigated 28 portals and 70,000 sites. A single AI hub now instantly surfaces answers, lifting self-service rates 87%.
Lab supply orders were handwritten in notebooks. Digital ordering now takes seconds, saving 30,000 hours for research annually.
Experts spent 15 minutes pulling data from scattered systems. Natural language prompts now generate detailed reports instantly.