I'm a software developer. I enjoy building things with Python, TypeScript (or JavaScript), Deep-/Machine Learning, backend APIs and AWS CDK.
Here I try to document small isolated solutions or processes I came across during work.
A short documentation on teacher forcing to training recurrent neural networks. Additional a scheduled sampling module to control teacher forcing prcoess during training.
An self-educational documentation on how end-to-end system in sound classification work. The data domain are environmental sounds.
An introduction to pre-commit and how to build custom python hooks for pre-commit.
A short guide on how to secure a cloudfront distribution using signed cookies, based on CF鈥檚 trusted key groups (including aws-cdk infra).
A simple solution to secure cloudfront distributions via lambda@edge
Tooling for a conviniently lightweight python project, using pyproject.toml with some simple tools
Mocking aws services and functionality with moto and pytest.
Some awesome tips and tricks one can do with pydantic. Data validation with pydantic and never load json without pydantic anymore. Example usage with boto3 (aws sdk for python).
An isolated example to show github authorization-code oauth flow in fastapi for web application flow + simple HttpBearer route dependency. This example can be used for all providers of code-based oauth authorization patterns.
A guide to understand how pretraining transformer models work. It鈥檚 basically the prestage, before usage with Huggingface is possible. See the full guide at https://github.com/arrrrrmin/albert-guide.