Netflix Engineer Open-Sources Headroom, a Token Compression Layer Claiming 95% Savings on JSON Data

Tejas Chopra, a senior engineer at Netflix, has released Headroom as open source—a compression proxy that strips redundant data before it reaches large language models, claiming to reduce token consumption by 60-95% on structured data and 20% on coding agents while maintaining identical model outputs. The project has already accumulated over 30,000 GitHub stars since January 2026 and reportedly saved users approximately $700,000 collectively while freeing 200 billion tokens for redeployment.