Automated AI code review that runs on every commit - repo fine-tuned suggestions

Executive Summary

Code review is slow and inconsistent. Provide an AI reviewer that runs on every commit/PR, using pre-trained models then fine-tuning on the repo to produce context-aware, private suggestions and linting. Large open models plus efficient fine-tuning techniques make it feasible to adapt models to a single repo at low compute cost, as the source highlights with the two-step pre-training and discriminative fine-tuning approach. At the same time, CI/CD adoption and Git-based workflows mean reviews happen frequently - every commit or PR - creating high cadence usage that justifies a recurring SaaS or per-seat model. Rising concern about sending proprietary code to external APIs also creates demand for on-prem or repo-local fine-tuning, which the source author aims to support. Use a two-step approach described in the source - generative pre-training for broad code knowledge then discriminative fine-tuning on the target repo - to deliver repo-specific, context-aware reviews that run inside CI or on-prem. Source evidence: the dev.to post by Shrijith Venkatramana explicitly describes building git-lrc and calls out generative pre-training plus discriminative fine-tuning as the recipe, implying a hybrid global model plus repo-specific tuning which creates a technical and privacy advantage when run in repo CI.