Iโm a self-taught AI enthusiast exploring Deep Learning and AI in depth, building and experimenting with models, algorithms and end-to-end systems, all built from scratch.
-
๐ Deep Learning Framework (repo)
- From scratch with NumPy
- Autodiff engine
- Modular API
- Implementations of CNNs, LSTMs, FNNs
- Toy demos like Shakespeare, MNIST or Titanic
-
๐ง Transformer Variants (repo)
- Encoder-only Transformer
- Convolutional low-rank approximation of attention
- Improvement of stability and efficiency
- Extended abstract redaction
-
๐ฎ NeuroEvolution (repo)
- NEAT framework
- Experiments with toy environments
-
๐ iXBRL and LLMs (research project in development)
- Parsing of financial filings
- Dataset creation
- Benchmarking
- Self-supervised learning
- Financial analysis