I led the development of a fully custom, multi-track live video editor built entirely in-house — including its timeline engine, video canvas, and real-time collaboration layer. Unlike conventional editors that depend on heavy third-party SDKs, this was built from the ground up using React, Canvas, and FFmpeg, designed for both real-time editing and server-side rendering through AWS Lambda.
Tech Stack
React, Canvas, Liveblocks, AWS Lambda
My Role
The system converts both user queries and video transcripts into dense vector embeddings using a sentence-transformer model. These embeddings are indexed in ElasticSearch, enabling semantic retrieval based on similarity scores rather than keyword matching.
1. Built the timeline system for arranging and trimming multiple tracks (video, audio, text, and overlays)
2. Designed and implemented the Canvas rendering engine, handling animations, transitions, and keyframes
3. Integrated React DnD Kit for intuitive drag-and-drop clip management
4. Developed the Lambda-based FFmpeg rendering pipeline for exporting final videos directly from the browser interface
Impact
1. Built the foundation for all subsequent in-house editing tools.
2. Enabled the team to produce multi-layered, animated content without external editors like Premiere or After Effects.
3. Achieved real-time collaborative editing for the first time in the company’s workflow