Tejas Chopra created Headroom and co-founded Headroom Labs after working on infrastructure at Netflix. Headroom reduces the context sent to language models while retaining originals that an agent can retrieve when needed.[1]
Start with Headroom →
Career
At Netflix, Chopra worked on the data-storage platform supporting studio and streaming workloads. His conference biography also describes earlier storage-infrastructure work at Box and a Carnegie Mellon master’s in electrical and computer engineering focused on computer systems. In a July 2026 interview, he says he left Netflix to build Headroom Labs full time.[4][1]
Contributions
Headroom sits between an application or agent and its model provider. It routes JSON, source code and prose to different compressors, removing repetition before the model receives the context. A local cache keeps the original material, and a retrieval tool lets the model ask for it back. The repository provides a library, local proxy, agent wrapper and MCP server. This architecture tries to lower repeated reading costs without making every compression decision permanent. The README cautions that savings depend on the workload: repetitive logs and arrays compress more than dense prose.[3]
Tools, writing & talks
Tools
Start here
Headroom
github.comLocal context compression and retrieval for agents, offered through a library, proxy, wrappers and MCP tools.
Read Headroom’s architecture overview, then run its own savings measurement on representative traffic. You need an existing agent or LLM application; the important first check is whether compressed context still supports your actual task.
Creator; Headroom Labs repository[3]Original source
Writing, research & talks
- How Headroom went from side project to enterprise infrastructureaiacceleratorinstitute.com
A first-person interview covering reversible compression, failure cases and the move from Netflix into Headroom Labs.
Interview subject[1]