Peter Song

Peter Song is a machine learning, data engineer, and AI writer with a focus on practical ML workflows. He shares insights on machine learning, data tools, and MLOps.

Embeddings vs One-Hot Tradeoffs: Making the Right Choice for Categorical Data

by

When working with categorical data in machine learning, one of the most consequential decisions you’ll make is how to represent these variables numerically. Two dominant approaches—one-hot encoding and embeddings—offer vastly different trade-offs in terms of dimensionality, computational efficiency, semantic representation, and model performance. While one-hot encoding has served as the traditional go-to method for decades, … Read more

Optimizing Embedding Generation Throughput for Large Document Stores

by

When you’re sitting on a corpus of 10 million documents and need to generate embeddings for vector search, semantic analysis, or RAG systems, raw throughput becomes your primary concern. A naive implementation processing documents one at a time might take weeks to complete, consuming compute resources inefficiently and delaying your project timeline. Optimizing embedding generation … Read more

Batch Inference Examples in AI

by

While real-time inference captures headlines with its instant predictions and interactive experiences, batch inference quietly powers some of the most impactful AI applications in production today. From Netflix generating personalized recommendations for millions of users overnight to financial institutions scoring credit risk across entire portfolios, batch inference enables AI systems to process massive datasets efficiently … Read more

Scaling vs Standardization: Choosing the Right Feature Transformation

by

In the realm of machine learning preprocessing, few decisions are as fundamental yet frequently misunderstood as choosing between scaling and standardization. These two feature transformation techniques appear similar at first glance—both modify the range and distribution of numerical features—but they operate through distinctly different mathematical mechanisms and produce results with profoundly different properties. The choice … Read more

Gradient Computation in Deep Learning: The Engine Behind Neural Network Training

by

Every time a neural network learns to recognize a face, translate a sentence, or predict stock prices, gradient computation is working behind the scenes. This fundamental mechanism is what transforms a randomly initialized network into a powerful prediction machine. Understanding gradient computation isn’t just an academic exercise—it’s the key to comprehending how deep learning actually … Read more

User-Based Collaborative Filtering Example

by

Recommendation systems have become an integral part of our digital experience, from Netflix suggesting your next binge-worthy series to Amazon recommending products you might love. At the heart of many of these systems lies a powerful technique called user-based collaborative filtering. In this comprehensive guide, we’ll dive deep into a practical user-based collaborative filtering example, … Read more

Cursor AI for Feature Engineering

by

When you’re developing machine learning models, feature engineering often consumes more time than model training itself—identifying relevant features, creating transformations, handling missing values, encoding categorical variables, and iterating through countless combinations to find what works. Cursor AI, the AI-native code editor, is transforming this traditionally manual and expertise-intensive process by bringing intelligent assistance directly into … Read more

Feature Scaling Issues with Tree-Based vs Linear Models

by

One of the most fundamental decisions in machine learning preprocessing is whether to apply feature scaling to your dataset. This seemingly straightforward choice has profound implications for model performance, yet it’s frequently misunderstood or applied inconsistently. The crux of the matter lies in understanding how different model families process numerical features—specifically, the stark contrast between … Read more

Positional Encoding Types in Transformers

by

The transformer architecture revolutionized natural language processing and has since expanded to dominate computer vision, speech recognition, and numerous other domains. At the heart of this architecture lies a crucial but often misunderstood component: positional encoding. Unlike recurrent neural networks that process sequences step by step, transformers process entire sequences simultaneously through self-attention mechanisms. This … Read more

Best Practices for Joining Large Fact Tables for ML Training Sets

by

Creating machine learning training datasets from production data warehouses is a deceptively complex challenge. While the conceptual task seems straightforward—join relevant tables to create a wide feature matrix—the reality involves navigating massive fact tables with billions of rows, managing complex join conditions that create fan-outs, balancing computational resources, and ensuring temporal consistency that prevents label … Read more