What You Can Explore
Working demos, evaluation results and implementation notes show how each system behaves and where its limits are.
AI SystemsResearch and operational workflows with source checks and approval steps.
Model EvaluationTask-level comparisons of accuracy, speed and cost, followed by checks after deployment.
Measurement SystemsRisk scoring, customer analysis and ranking models with documented evaluation.
How the Systems Are Built Architecture, deployment and evaluation
Built with
Claude Codeagent in GitHub Actions
Anthropic APIClaude models
AWSLambda · S3 · Bedrock
GitHub Actionstests · deploys · checks
Vercelsite and functions
Postgreson Neon
Implementation
From repository to production
1 · GitHub
- Automated tests and proposed changes
- Human review before merge
2 · Deploy
- Vercel for the site and functions
- AWS Lambda and S3 for models
3 · Model Calls
- Anthropic API and AWS Bedrock
- Token use and cost recorded
4 · Data
- Read-only tools and SQL allow-lists
- Citation checks on research answers
5 · Checks
- Fixed reference inputs every six hours
- Public run history →
Contributors
Project-based collaborationThe Bench
Senior Engineers
Experience at Amazon, Oracle and HubSpot. Contributors to scoped builds, technical reviews and production hardening.
Industry Specialists
Domain expertise for research, workflow design and review, brought in where the project needs it.