Missing data is a major headache in data science, undermining the reliability of analyses. Think of it like trying to solve a puzzle with missing pieces – frustrating, right? Traditional methods for filling these gaps often fall short, especially when dealing with "derived" missing values that require complex calculations based on other data points. Imagine trying to calculate total sales with a missing quantity value – you need more than just an average to get the right answer. Researchers are now turning to Large Language Models (LLMs), the brains behind AI chatbots, to tackle this challenge. But how do you teach an LLM to reason like a data scientist? A new technique called SketchFill offers a clever solution. It provides the LLM with a "sketch," a set of high-level instructions similar to how a data expert would approach the problem. This sketch guides the LLM to generate code that accurately calculates the missing values, even when complex formulas are involved. It’s like giving the LLM a cheat sheet for solving data puzzles. SketchFill also incorporates a "reflector" that checks the LLM's work, iteratively refining the code until it gets the right answer. In tests across various datasets, from financial markets to supermarket sales, SketchFill significantly outperformed existing methods, boasting up to 74% accuracy in some cases. This represents a major leap forward in automated data cleaning and holds the promise of dramatically speeding up data analysis workflows. While challenges remain, such as dealing with noisy data or extremely complex calculations, SketchFill opens exciting new avenues for leveraging the power of LLMs in the fight against missing data. It’s a big step towards letting AI do the tedious work, freeing up data scientists to focus on the bigger picture.