How Chroma cut average review time by 40+ hours/PR

Challenge

As Chroma's engineering team grew, review cycles became a bottleneck and was slowing down development. Growth brought more code, more contributors, and longer review cycles for senior engineers. Valuable engineering time was being spent enforcing standards instead of driving new initiatives.

Approach

Chroma sought to maintain its high code quality while scaling contributor volume. The team integrated Propel's AI code reviewer directly into their GitHub workflow to provide precise, context-aware feedback on every pull request.

Each comment was specific, meaningful, and explainable, helping developers understand why something needed to change rather than just what to fix. This allowed Chroma to improve review depth and accuracy without increasing manual workload or adding reviewers.

Solution

Propel's reviewer fit seamlessly into existing workflows, pairing contextual suggestions with clear rationales so the team could move from observation to action fast.

• Feedback arrives automatically alongside every PR.
• Guidance references schemas, protocols, and internal patterns that matter most.
• Engineers stay in flow with concise, high-signal comments ready to action.

Results

Propel became an active reviewer within Chroma's team, reinforcing standards, speeding up collaboration, and ensuring every PR moves the codebase forward with intent. Propel became a trusted set of eyes across every pull request, reinforcing standards and surfacing actionable improvements before anyone had to even look.

Chroma’s engineers could move faster with confidence, knowing every PR received a consistent, high-quality review.

• 40+ hour reduction on average PR review time.
• Higher review accuracy across critical backend services.
• More consistent engineering standards without adding headcount.