A real-time platform for live moments

The commercial value of live sport lives in moments — the goal, the penalty, the knockout — and those moments are usually missed. Fans split their attention across work, travel and several matches at once, and by the time a highlight package lands, the excitement has passed. HypeDirector closes that gap in seconds, and Toobler engineered the platform that makes it possible: AI-and-human moment detection, an event-driven cloud backend, and purpose-built apps for everyone who touches a live event.

Customer and context

HypeDirector is a real-time sports-moment intelligence platform: it captures, processes and distributes live event highlights within seconds of them happening. It is built for professional sports organizations, broadcasters, publishers, esports companies and media partners, and it serves five distinct user groups, each through its own application:

• Fans — personalized live notifications, instant highlight clips, saved moments and interactive feeds.
• Content Owners — a console to run live sessions, publisher profiles, audience engagement, social publishing, moderation and monetization.
• Spotters — the editorial layer who identify and validate live moments through dedicated live-ops workflows.
• Operations Teams — moderation, onboarding, taxonomy, notifications, governance and platform health.
• Technology Partners — secure, programmatic access to live moment data through enterprise APIs and webhooks.

Toobler is the engineering partner that built the platform end to end — the applications, the AI and the event-driven cloud infrastructure underneath them.

The challenge

A real-time sports-moment platform is a hard problem on every axis at once:

• Fan engagement decays fast. Traditional workflows are built around full-event viewing and post-match highlights. The value is in the live moment, and it has to reach the right fan while the excitement is still there.
• Publishers don’t own their audience. Relying on third-party social and broadcast platforms makes it hard to build direct relationships, collect first-party data, and create recurring digital revenue.
• Live operations are complex. Capturing moments at scale means coordinating editorial teams, scheduling spotters, moderating content, managing notifications and maintaining quality across many publishers simultaneously.
• Traffic is spiky and unforgiving. Thousands of live notifications, personalized feeds, generated clips and partner calls have to hold up during a high-profile event, then scale back down.
• It had to become a platform, not just an app. As demand for live sports data grew, HypeDirector needed to become reusable infrastructure that external organizations could integrate into their own products.

The solution

Toobler built HypeDirector as a multi-sided, AI-powered platform where each stakeholder works through a purpose-built application over a shared, event-driven backend. Rather than pushing every live moment through one monolithic workflow, responsibilities are split across independent product layers and cloud services, so the platform scales without putting the live experience at risk.

One event model, many outcomes. When an event begins, moments originate from trained editorial spotters or from AI detection. Every detected moment is standardized into a common data model, then routed through asynchronous cloud workflows that — depending on the moment type — act in parallel to:

• notify subscribed fans;
• update personalized feeds;
• generate video clips;
• publish to connected social channels;
• deliver to external partners;
• record analytics.

Personalized fan experiences. The fan app replaces scrubbing through broadcasts with notifications based on the teams, competitions and publishers each fan follows — discover live sessions, get instant moment alerts, watch auto-generated clips, and build a personal highlight library they can save, like, comment on and share.

A console for content owners. Owners run every part of live-event management from one place: creating and managing live sessions, configuring publisher profiles, overseeing engagement, controlling social publishing, setting moderation policy, and preparing monetization — direct ownership of the audience relationship rather than renting it from other platforms.

Editorial operations that scale. Dedicated spotter workflows let trained operators capture moments with speed and precision — shift assignments, moment signaling, activity review, performance monitoring and payment workflows — a human-in-the-loop model that keeps editorial confidence high alongside the automation.

AI as a core capability. Machine-learning services process live and archived media continuously — speech transcription, event classification, rolling video analysis and automated clip generation — using Google Speech-to-Text, Vertex AI and Gemini models, BigQuery, and MediaMTX. AI complements the human spotters, widening event coverage while reducing manual effort.

An event-driven backend for reliable distribution. Built on Firebase Cloud Functions, Google Cloud, Pub/Sub, Cloud Tasks, Firestore, BigQuery, Secret Manager and Terraform, the platform processes notifications, feed updates, clip generation and partner integrations asynchronously. Heavy work is spread across resilient cloud workers with retry-safe processing, deterministic identifiers, pagination and resumable jobs — which holds reliability up during peak activity and keeps duplicate processing and timeouts down.

A Moment Platform for partners. HypeDirector evolved into reusable media infrastructure. Its Moment Platform gives approved organizations secure access, through REST APIs and signed webhooks, to:

• sessions and live moments;
• subscriptions and webhook events;
• delivery logs and replay;
• test environments.

Broadcasters, clubs, sponsors, vendors and analytics providers can integrate without touching internal cloud infrastructure.

Social-first amplification and governance. Eligible clips are published to owner-connected social channels by dedicated workers, asynchronously, so social distribution never interrupts live processing. A central operations toolset handles onboarding, moderation, taxonomy, notification policy, publisher approvals, spotter management and platform health.

Engineering worth naming

• The event pipeline is the product. One standardized moment model fanning out — notify, feed, clip, social, partner, analytics — asynchronously and in parallel is what lets the platform serve every audience at once without disrupting the live event.
• AI and humans in one loop. Vertex AI, Gemini and Speech-to-Text detection working alongside a trained spotter layer — coverage at machine scale, with editorial confidence.
• Built to survive a spike. Retry-safe, deterministic, resumable cloud workers on Pub/Sub and Cloud Tasks — engineered for the traffic pattern of a live final, not an average day.
• A product that became a platform. The same live-moment core exposed as secure REST APIs and signed webhooks, turning moments into reusable business assets.

Results

• Live moments reach the right fans in seconds — detected, standardized, clipped and delivered while the excitement is still live, personalized to what each fan follows.
• Five stakeholder experiences on one backend — fans, content owners, spotters, operations and technology partners, each with purpose-built tools.
• Publishers own their audience — direct engagement, first-party relationships and monetization, instead of depending on third-party platforms.
• A reusable media platform — the Moment Platform lets external organizations integrate live sports data securely, extending the product well beyond its own apps.
• An architecture built for AI and growth — event-driven, cloud-native, and extensible toward further AI automation and commercial services.

What this means for real-time and live-media product engineering

Real-time media is unforgiving: the value is entirely in the seconds between something happening and the right person knowing, and it has to hold up when everyone shows up at once. HypeDirector is that problem solved end to end — AI-and-human moment detection, an event-driven cloud backend engineered for spikes, five purpose-built applications, and a partner platform on top — built by a single engineering partner. If you are building a real-time or live-media product and need the applications, the AI, and the infrastructure that keeps it fast under load, talk to Toobler’s engineering team.