Inside Meitu’s 6th Tech Salon: Deep Dive into Meipai’s Recommendation, Monitoring, and Live‑Streaming Architecture

On September 23, Meitu held the sixth “Meipai Technology Decoding” salon at Beijing Water Cube, attracting about 140 technical elites from nearly 400 registrations.

Key Speakers and Topics

Data Director Zhu Yi presented the evolution of Meipai’s recommendation system, discussing early implementations, challenges faced, and optimization strategies that leveraged massive user data to improve playback and click‑through rates.

Architecture Senior Manager Wang Jingbo explained the three‑stage evolution of Meipai’s live‑stream bullet‑screen system, detailing how the platform maintains stability and high interaction rates while supporting millions of concurrent users.

Operations Director Wang Guansheng described Meipai’s comprehensive monitoring framework, addressing alarm storms, improving alert efficiency, and sharing industry‑wide monitoring case studies.

Architecture Director Zhu Lei analyzed common live‑stream architectures, compared their pros and cons, and highlighted Meipai’s specific optimizations for user experience and performance.

Technical Highlights

Meipai’s “Hundred‑Change Background” effect originates from MTLAB’s real‑time background segmentation, enabling fast background detection, replacement, and AR effects.

The platform’s live‑stream infrastructure now supports hundreds of thousands of simultaneous viewers, with notable events such as TFBOYS’ first live broadcast reaching over one million participants.

Event Atmosphere

Attendees engaged in Q&A sessions, sharing experiences and discussing future directions in computer vision, big data, and AI‑driven video services.

Meitu plans to continue deepening its research in computer vision, leveraging its massive data assets to rapidly integrate cutting‑edge technologies into products and enhance user experience.