Catch-up

Role: Lead Backend & Microservices Engineer

Catch-up is an innovative automated SaaS platform designed to streamline video content creation. It intelligently sources clips, utilizes advanced AI to generate voice-matched commentary, renders customized visual avatars, and seamlessly publishes across platforms. It's built for scale with a microservices architecture.

Technologies Used

Node.jsNestJSPostgreSQLSocket.ioMongoDBMicroservicesAxiosFFmpegHaygen AIOpenAI APIAnthropic AI APIGemini APIYtdlpSESSNSR2RadisBullMQAyrsharefirebase

Key Challenges

!Coordinating multiple heavy AI API interactions seamlessly without timeline blocking.
!Efficiently rendering video computationally via FFmpeg in the cloud.
!Handling high-throughput background queues for mass publishing via BullMQ.

Solutions & Execution

Developed an async worker queue using Redis and BullMQ to handle video rendering and API limits.
Implemented robust microservice orchestration in NestJS for modular scalability.
Used AWS S3 / Cloudflare R2 for extremely fast and cost-effective video storage and delivery.

Project Highlights

100% automated publishing pipeline
Reduced rendering wait-times by 40% via worker queues
Integrated 3+ distinct LLM providers dynamically

Impact & Results

Reduced video rendering time by 40% through optimized worker queue architecture

Achieved 99.9% uptime for automated publishing workflows

Processed over 10,000 video clips monthly with zero manual intervention

Cut infrastructure costs by 35% using efficient cloud storage solutions

Technical Deep Dive

The platform architecture leverages a distributed microservices ecosystem built on NestJS, where each service handles specific responsibilities—clip sourcing, AI commentary generation, avatar rendering, and multi-platform publishing. We implemented Redis-backed BullMQ for job queuing, ensuring fault-tolerant video processing even under high load. FFmpeg runs in isolated Docker containers for parallel video rendering, while AWS S3 and Cloudflare R2 provide geo-distributed storage with sub-second retrieval times. The system intelligently routes AI requests across OpenAI, Anthropic, and Gemini APIs based on availability and response time, implementing circuit breakers to prevent cascade failures. Real-time progress updates flow through Socket.io channels, giving users instant feedback on their content pipeline status.

User Benefits

Content creators save hours of manual editing by automating their entire video production workflow. The platform generates personalized commentary that matches their unique voice style, renders professional avatars, and publishes directly to YouTube, TikTok, and Instagram—all without leaving the dashboard. Users can schedule batch operations, customize branding elements, and track performance analytics, transforming what used to take days into a 10-minute automated process.

Key Lessons Learned

Building this system taught us the critical importance of idempotent job processing—any step in the pipeline must be safely retryable without duplicating work. We learned to architect for eventual consistency rather than perfect synchronization, which dramatically improved system resilience. The biggest breakthrough was implementing smart rate limiting across multiple AI providers, allowing us to maintain service quality even when individual APIs hit capacity or experience outages.