Building an AI résumé scorer with Claude + AWS Bedrock - architecture and lessons learned

Short summary

ScorAxis is an AI résumé scorer that gives applicants a 0–100 ATS compatibility score with flagged issues and actionable fixes. The architecture splits scoring into deterministic Python logic (0–70) and Claude Sonnet via AWS Bedrock for AI checks (0–30), using forced tool_choice for reliable structured JSON output. Key lessons include separating deterministic from AI logic early and shipping fast for real feedback.

• •AI resume scorer using Claude Haiku for parsing and Sonnet for checks via AWS Bedrock
• •Scoring split: deterministic Python (0-70) + LLM tool-use (0-30)
• •Forced tool_choice yields near-zero JSON validation failures in production

Generated with AI, which can make mistakes.