work / 08 / Zero external calls

Private AI coding CLI on self-hosted Gemma 3

A terminal-native developer assistant on in-house GPUs - codebase-aware retrieval, tool use, multi-step execution - keeping AI-assisted development entirely inside the enterprise security perimeter.

1. 01 Terminal-native developer assistant powered by self-hosted Gemma 3 running on in-house GPUs.
2. 02 Codebase-aware context retrieval, tool use, and multi-step task execution: read, edit, run, verify.
3. 03 Zero external model calls - fully private AI-assisted development inside the enterprise security perimeter.
4. 04 Built on prior LLM work: fine-tuned open-source models on H100 GPUs, with a deployment pipeline onto GCP GPU clusters.