ShadowPersona

Inspiration

As a tech community developer lead in Northern Nigeria, I learned that organizing a technology event is never just choosing a date, booking a space, and publishing a flyer. In communities affected by insurgency and persistent insecurity, bringing people together means planning amid uncertainty, limited resources, and changing circumstances. Decisions about the event’s purpose, audience, format, partners, and budget matter because organizers have to make the best use of the time and resources their communities can commit.

Those experiences shaped the question behind ShadowPersona: how can an organizer make a more informed event plan when the signals they have are incomplete? A generic AI answer can sound confident, but confidence is not evidence—and a cultural match is not a promise that people will attend. I wanted to build a tool that helps organizers reason from real cultural signals and make their assumptions visible, while leaving local judgment where it belongs.

What we built

ShadowPersona is an event-planning intelligence layer. An organizer can provide a brief, cultural tags, or event details. The application resolves those cues through Qloo and uses Qloo’s returned artist and brand insights as the evidence for its analysis. Kimi K2.6 turns that evidence into a structured explanation and planning recommendations. Separately, the app calculates low, expected, and high event scenarios from assumptions entered by the organizer, such as local aggregate reach, conversion rates, ticket prices, capacity, and costs.

That separation is intentional: Qloo provides cultural-affinity evidence; organizer inputs drive the financial scenarios. Neither is presented as proof of attendance, purchase intent, or event safety. Optional public research can add cited context, but the core workflow does not need a personal profile or an attendee’s private history.

How we built it and what we learned

We built the application with Next.js and TypeScript, connected to Qloo’s API and Kimi’s structured-output workflow. We use schema validation at the boundaries and calculate event economics in application code rather than asking a language model to invent financial results. Supabase supports shared request admission, duplicate-run suppression, and private result caching.

One of our hardest design challenges was deciding where AI should stop. A model can explain why returned cultural evidence may be relevant, but it should not fabricate Qloo matches or turn affinity into a demand forecast. We therefore made the evidence visible, label assumptions, validate generated output, and keep forecasts as scenarios that organizers can inspect and change. We learned that a useful planning tool should expose uncertainty instead of hiding it behind polished language.

Where we want to go

ShadowPersona’s long-term direction is an operating system for event management, planning, and forecasting—not just a recommendation screen. The current product is an initial intelligence layer. Over time, we want to connect the event lifecycle around one organizer-owned plan: audience and cultural-fit exploration, format and market comparisons, venue and budget scenarios, partner discovery, timelines and run-of-show planning, operational checklists, and post-event learning.

The goal is not to automate away the knowledge of community organizers. It is to give them a clearer starting point, show which decisions rely on evidence versus assumptions, and help them adapt a plan as conditions change. Forecasts will remain estimates, and safety decisions will remain human-led; software cannot resolve insecurity or guarantee an outcome.

Why Qloo matters

Qloo is central to the product, not an ornamental API call. Its Taste AI gives ShadowPersona a way to connect cultural cues to real entities and cross-domain insights that a general-purpose language model could otherwise only guess at. The demo follows that chain: organizer context → resolved cultural tags → Qloo evidence → grounded recommendations → transparent, assumption-based scenarios.

That focused workflow is our strongest hackathon story: it shows a specific problem from my experience, a meaningful role for Qloo, a working end-to-end product, and a credible path toward a broader event operating system—without claiming that the current version already does everything.

Submission strategy (internal; optional, not part of the public story)

The strongest case for this project is to demonstrate the Qloo dependency clearly: show the same brief moving from tags to live Qloo evidence to recommendations, then show how changing organizer assumptions changes the deterministic forecast. Lead with the community-organizing problem, keep the demo narrow and reproducible, and be explicit about what is live versus planned. This makes the ambition memorable while keeping the claims credible; no story can guarantee a win.

Built With

• event
• event-management
• generative-ai
• kimi-k2.6
• next.js
• node.js
• postgresql
• qloo-api
• react
• rest-api
• supabase
• tailwind-css
• tavily-api
• typescript
• vercel
• zod