Developer Tools
Cross-Version ML Code Upgrade Tool
An AI-powered tool that modernizes legacy machine learning codebases by automatically detecting and upgrading deprecated library code. Uses Claude via OpenRouter API for intelligent code transformation with retry logic on failures. Features runtime validation that executes test commands after each upgrade to confirm functionality, with comprehensive markdown reports documenting all changes.
Tech Stack
7 technologies used
PythonOpenRouter APIClaudeTensorFlowPyTorchNumPyStreamlit
Key Features
- Automated detection and upgrading of deprecated ML library code (TensorFlow 1.x, PyTorch, NumPy)
- LLM-driven code transformation using Claude via OpenRouter with retry logic on failures
- Runtime validation mode executing test commands after each upgrade to verify functionality
- Dependency version updates with compatibility checking
- Markdown report generation documenting all code changes and transformations
- Customizable setup commands for environment preparation before upgrades
- Streamlit web interface for interactive usage and experimentation
- macOS archive artifact cleanup (automatic skipping of resource forks and metadata)
Challenges Solved
- Handling complex API changes that aren't simple 1:1 replacements across framework versions
- Maintaining semantic equivalence after code transformations
- Managing virtual environment creation and dependency resolution automatically
Outcomes & Impact
- Agentic upgrade pipeline: upload code → analyze → transform → validate → report
- Iterative refinement loop improves upgrade success rate on compilation/runtime errors
- Comprehensive diff reports showing all changes for review before deployment