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.

DMS Migration Strategies for Production Databases

by

Migrating production databases represents one of the most high-stakes operations in enterprise IT. Unlike test environments where failures are learning opportunities, production migrations must succeed while maintaining business continuity, preserving data integrity, and meeting strict uptime requirements. AWS Database Migration Service (DMS) has emerged as a powerful tool for these critical migrations, but simply spinning … Read more

Should I Run LLMs Locally?

by

The decision to run large language models locally versus using cloud-based APIs has become one of the most consequential technical choices facing developers and organizations today. As models have become more capable and accessible, the barriers to local deployment have lowered dramatically. Tools like Ollama, LM Studio, and llama.cpp make running sophisticated models on consumer … Read more

Exploring 15 Quirky Datasets for Creative Data Analysis

by

Data analysis doesn’t always have to involve corporate sales figures, customer demographics, or website traffic patterns. Some of the most engaging and enlightening analytical work happens when you explore unusual, quirky datasets that reveal unexpected patterns about the world around us. These unconventional datasets offer opportunities to practice analytical skills while discovering fascinating insights about … Read more

How to Build a Recommendation System with Minimal Code

by

Recommendation systems power some of the most successful products in technology—Netflix’s movie suggestions, Amazon’s product recommendations, Spotify’s playlists, and YouTube’s endless video queues. The sophistication of these systems might suggest they require extensive machine learning expertise and thousands of lines of code to implement. In reality, you can build surprisingly effective recommendation systems with remarkably … Read more

Data Storytelling Techniques for Non-Technical Audiences

by

The most brilliant analysis in the world becomes worthless if you cannot communicate its insights effectively. Data scientists, analysts, and technical professionals often fall into a common trap: they present data the way they understand it—through statistical measures, technical terminology, and complex visualizations—rather than the way their audience needs to receive it. When presenting to … Read more

Building Lightweight ETL Pipelines for Small Projects

by

Enterprise ETL tools like Informatica, Talend, and Apache Airflow are powerful but often overkill for small projects. When you’re building a startup MVP, automating internal reports, or aggregating data for a side project, you don’t need heavyweight infrastructure with dedicated servers, complex configuration, and steep learning curves. What you need is a lightweight ETL pipeline … Read more

Easy Ways to Optimise SQL Queries for Faster Performance

by

Slow SQL queries can cripple application performance, turning responsive user interfaces into frustrating experiences where users wait seconds or even minutes for data to load. The good news is that most performance problems stem from a handful of common issues that are relatively straightforward to fix once you understand what to look for. You don’t … Read more

Understanding Neural Networks with Real-World Examples

by

Neural networks have become the invisible infrastructure powering much of our digital lives, yet they remain mysterious to most people. When you unlock your phone with your face, ask Siri a question, or see personalized recommendations on Netflix, neural networks are working behind the scenes. The challenge is that explanations of neural networks typically fall … Read more

Best Way to Learn PyTorch: Strategic Approach to Mastering Deep Learning

by

PyTorch has emerged as the dominant framework for deep learning research and increasingly for production deployments. Its intuitive design, dynamic computation graphs, and Pythonic interface make it the preferred choice for both researchers pushing the boundaries of AI and engineers building practical machine learning systems. However, the path to PyTorch mastery is not always obvious, … Read more

Building a Retrieval Augmented Generation (RAG) Pipeline with LLM

by

Large Language Models have transformed how we interact with information, but they come with a significant limitation: their knowledge is frozen at the time of training. When you ask an LLM about recent events, proprietary company data, or specialized domain knowledge, it simply cannot provide accurate answers because it has never seen that information. This … Read more