What to Build
You can build anything, as long as it is powered by KumoRFM. Some ideas include:

• Dating Match Predictor - Estimate compatibility between users using relational profile, interaction, and behavioral data.

• Personalized Interaction Agent - An AI agent that uses KumoRFM predictions to adapt conversations, recommendations, or decisions in real time.

• Business Intelligence Dashboard with Predictive Insights - Predict sales, churn, or lead scores, and integrate outputs directly into BI tools like Tableau, Looker, or Power BI.

• Autonomous Fraud Detection & Response Agent - Detect suspicious transactions and trigger automated follow-up actions (alerts, account holds, extra verification).

• Dynamic Pricing Agent - Continuously adjust prices based on predicted demand, competitor pricing, and supply chain signals.

• Predictive Maintenance Orchestrator - Forecast equipment failures and automatically create work orders or service tickets via integration with ops tools.

• Personalized Learning or Onboarding Agent - Recommend the next best training or onboarding steps for users, adapting in real time as performance changes.

• Supply Chain Risk Mitigation Agent - Predict delays or stockouts and trigger automated reordering or supplier alerts.

• Healthcare Readmission Prevention Agent - Predict high-risk patients and prompt case managers with personalized intervention recommendations.

REGISTER HERE: https://lu.ma/w0xg3dct

Top Prizes

• 1st Place: $10,000

• 2nd Place: $7,000

• 3rd Place: $3,000

• In addition to cash prizes, winners will gain visibility through Kumo’s marketing channels and event coverage, and may be considered for sponsored opportunities, strategic partnerships, internships, or potential employment with Kumo or our industry partners.

Dataset Guidelines
You may use a public dataset (such as from Kaggle) or a mock dataset. All predictions must be generated using KumoRFM. You do not need to share any datasets with Kumo. All computations are performed in memory and used only for generating predictions during runtime.

Judging Criteria
Judges will include Jure Leskovec, Hema Raghavan, and other Kumo team members. Projects will be evaluated based on:

• Practical Impact (30 points) – Does the project address a real, meaningful problem with clear potential value in the real world?

• KumoRFM Implementation & Innovation (30 points) – How effectively and creatively does the solution leverage KumoRFM’s capabilities (e.g., relational data handling, predictive query language)? Is the integration technically sound and central to the solution?

• Technical Execution (20 points) – Is the solution functional, robust, and well-engineered? Does it demonstrate good architecture and performance?

• Storytelling & Presentation (20 points) – How clearly and compellingly does the team explain the problem, solution, and impact? Is the demo engaging and easy to follow?