Introduction
Aura Arena is an AI-powered social engagement platform designed around short-form video dance challenges. Creators and admins can host challenges requiring users to replicate specific dance steps or routines directly against an original video. By combining social feed mechanics (similar to Instagram) with automated AI scoring, Aura Arena creates a gamified experience where users submit dance attempts and receive instant feedback and scores.About the Client
The client aimed to build a high-engagement, MVP-first social video platform that connects creators, users, and brands.- Core Concept: Creators and admins post dance challenge videos. Users submit replicated dance videos to compete.
- Monetization & Brand Model: Brands can sponsor challenges and pay the platform to run campaign video contests to drive user engagement and ROI.
- Vision: Deliver an intuitive, Instagram-like social feed with seamless video upload, automated video scoring, payment integrations, and deep admin analytics.
Business Impact
- Rapid MVP Deployment: Successfully delivered a feature-complete MVP under tight timelines by decoupling app services and leveraging lightweight streaming pipelines.
- New Revenue Streams: Enabled brand-sponsored campaigns with integrated payment systems.
- High Scalability: Built a media and AI pipeline capable of smoothly handling thousands of concurrent user uploads without crashing feed performance or server infrastructure.
- Gamified User Engagement: Automated AI scoring and instant scorecards created a competitive environment driving high video submission rates.
Challenges
Technical & AI Challenges- Server Overload on High Upload Volumes: Managing ~10,000 video uploads simultaneously created severe backend congestion and server bottlenecks.
- Bandwidth & Latency: High-bitrate videos (40–50 Mbps) caused slow feed load times and excessive bandwidth consumption.
- AI Dance Comparison & Scoring: Comparing user video submissions against original videos using Gemini required balancing AI prompt token limits, cost, and video frame synchronization.
- Simultaneous Processing: Handling video compression, 3D/pose recognition, and AI scoring at the same time without delaying user feedback.
- Admin & Creator Analytics: Admins and creators required real-time visibility into video performance, scoring metrics, and campaign engagement.
- Platform Consistency: Ensuring UI/UX consistency across the Flutter mobile app and React-based web admin portal.
Metrics
| Metric / Dimension | Before Solution (Baseline) | Implementation Strategy | After Solution (Results) |
| Video Bitrate & File Size | High raw video uploads at 40–50 Mbps, consuming massive bandwidth. | Integrated AWS MediaConvert pipeline to process and compress videos into HLS / MP4 formats. | Dynamic HLS adaptive streaming with minimal bandwidth usage and no loss in video quality. |
| Video Feed Latency | Slow feed response and video buffering due to raw file playback. | Implemented bit-by-bit chunked streaming and decoupled media storage from the app backend. | Fast, lag-free Instagram-like feed scrolling and instant video playback. |
| Server & Bulk Processing | High risk of server overload and crashes during ~10,000 video upload peaks. | Decoupled core backend APIs from background video transcoding and AI processing jobs. | Smoothly handles peak challenge traffic of ~10,000 uploads without server congestion. |
| AI Evaluation & Cost | High token consumption and costs due to uncompressed prompts and long context. | Compressed text context and structured prompt injection strategies for Gemini AI. | Optimized token efficiency, lower AI operational costs, and higher scoring accuracy. |
| Dance Scoring Sync | Delay and misalignment between dance step detection and score generation. | Synchronized video keyframes directly with Gemini AI pose and step analysis logic. | Real-time scorecards generated immediately upon user video submission. |
| Admin & Brand Analytics | Limited visibility into campaign performance and submission tracking. | Developed dedicated analytics and campaign management modules on the React.js web portal. | Comprehensive real-time dashboard tracking challenge engagement, ROI, and user scores. |
Tech Stack
| Domain | Technology / Tool |
| Mobile App | Flutter (Android / iOS) |
| Web / Admin Panel | React.js |
| Backend Framework | Node.js, Express.js |
| Database | MongoDB |
| AI Model | Google Gemini AI (Dance & Pose Step Analysis) |
| Media & Cloud Pipeline | AWS MediaConvert, HLS Streaming, MP4 Processing |
| Integrations | Payment Gateways, Media Storage |
- 1 x Mobile App Developer (Flutter)
- 1 x Frontend Developer (React.js Admin Panel)
- 1 x Backend Developer (Node.js/Express/MongoDB)
- 1 x AI Developer (Gemini Integration & Prompt Engineering)
- 1 x Business Analyst / Project Manager
Our Approach
- Decoupled System Architecture: To prevent server crashes during peak challenges, we decoupled the core mobile app API from the heavy media transcoding and AI scoring jobs.
- Optimized AWS Media Pipeline: Converted raw 40–50 Mbps video uploads into streamable HLS & optimized MP4 formats via AWS MediaConvert. This ensured bit-by-bit stream processing without loss of visual quality.
- Smart AI Scoring Engine (Gemini AI): Designed a specialized prompt injection workflow. By compressing context tokens and feeding synced keyframes, Gemini compares original dance steps against user submissions to generate dynamic scorecards.
- Instagram-Inspired UI/UX: Built a streamlined, short-form video feed user journey (Upload → Process → Score → Share) designed specifically to feel familiar and fast for users.
- Creator & Brand Monetization Suite: Integrated custom campaign modules and web dashboards for brands and admins to run paid video challenges and track real-time analytics.
