Philip Sarajlic — Machine Learning Engineering & Applied Data Science
Explaining Complex Machine Learning Models with SHAP
How to Deploy Quantized Large Language Models (LLMs) Locally
K-means Clustering: A Practical Guide to Understanding Customer Patterns
Evaluating the Trade-Off Between Performance and Complexity in L1-Penalized Logistic Regression
Beyond CPU Boundaries: How to Harness GPUs for Neural Network
Why GPUs accelerate neural-network training and hyperparameter tuning, how memory bandwidth sets the limit, and how to move beyond CPU boundaries.
Extending scikit-learn model functionality through inheritance
How to extend scikit-learn by inheriting its base classes, building custom estimators that work in pipelines and honor the fit/predict contract.
Dimensionality Reduction with Principal Component Analysis (PCA)
How PCA reduces dimensions, why standardizing features first matters, and how to read explained variance, with worked Python examples.
K-Medoids Clustering: A More Robust Alternative
How medoids resist outliers, how K-medoids (PAM) differs from K-means, and when to choose it, with worked Python examples.
Dimensionality reduction with Uniform Manifold Approximation and Projection (UMAP)
How UMAP reduces high-dimensional data to a 2D map, when it beats PCA, and how to tune its key parameters, with interactive visuals and Python.
Using Docker to Access Linux-Only ML Training Tools from Windows
How Docker runs Linux-only machine-learning tools from Windows, what containers actually do, and how to reach the GPU through WSL2.
Accelerating Inference: MLX vs. GGUF Models Compared
How MLX and GGUF differ on speed and time-to-first-token for local LLM inference, and which to choose on Apple Silicon, with benchmarks.
Unlocking Massive LLM Potential with Unified Memory
How Apple Silicon’s unified memory lets one machine load far larger language models, why memory placement matters, and what it means for local LLMs.
How to Perform Correlation Analysis
How to measure and interpret correlation in Python, how Pearson and Spearman differ, how to read a correlation matrix, and the common pitfalls.
Trending Articles
Extending scikit-learn model functionality through inheritance
How to extend scikit-learn by inheriting its base classes, building custom estimators that work in pipelines and honor the fit/predict…
Dimensionality Reduction with Principal Component Analysis (PCA)
How PCA reduces dimensions, why standardizing features first matters, and how to read explained variance, with worked Python examples.
Beyond CPU Boundaries: How to Harness GPUs for Neural Network Tuning
Why GPUs accelerate neural-network training and hyperparameter tuning, how memory bandwidth sets the limit, and how to move beyond CPU…
K-means Clustering: A Practical Guide to Understanding Customer Patterns
How K-means groups data, why feature scaling matters, and how to find customer segments in Python, with a hands-on worked…
Accelerating Inference: MLX vs. GGUF Models Compared
How MLX and GGUF differ on speed and time-to-first-token for local LLM inference, and which to choose on Apple Silicon,…
