I like understanding how things work under the hood before reaching for a framework, from hand-deriving backprop gradients to reproducing studies from scratch.
- 🔧 At the intersection of ML engineering and software systems
- 🧪 Research interests: mechanistic interpretability, LLM evaluation, time-series forecasting
- 📄 I read papers and build them - not just use the library
- 🚀 Always shipping
|
9-model ensemble (statistical, tree-based, neural) forecasting order volumes with 101-fold CV. Hampel filtering & Fourier engineering. |
Reproducing SOTA papers on LLM coding. Benchmarks different LangChain chains with Ragas-based significance & cognitive complexity scoring. |
|
Retrieval-augmented search with vector retrieval & client-side citation matching with in-browser ONNX embeddings. |
2-stage transformer pipeline: trend classifier & magnitude regressor. Custom feature engineering with momentum & seasonality. |