BVA Data Factory’s Post

“Data science is probably not dead, but surely it is changing. The best data scientist will not be who can code faster, but who can better direct the assembling of the data science project taking into account data integration, data quality, data history, machine learning algorithms, result interpretation, and correctness of the process.“ Read this story from Rosaria Silipo on Medium: https://lnkd.in/gHd26YcV

“Data science is probably not dead, but surely it is changing. The best data scientist will not be who can code faster, but who can better direct the assembling of the data science project taking into account data integration, data quality, data history, machine learning algorithms, result interpretation, and correctness of the process.“ Read this story from Rosaria Silipo on Medium: https://lnkd.in/euA5cGmZ

“Data science is probably not dead, but surely it is changing. The best data scientist will not be who can code faster, but who can better direct the assembling of the data science project taking into account data integration, data quality, data history, machine learning algorithms, result interpretation, and correctness of the process.“ Read this story from Rosaria Silipo on Medium: https://lnkd.in/dTGG5BEy

DSBENCH: How Far Are Data Science Agents to Becoming Data Science Experts? The research paper introduces a new benchmark, DSBench, designed to evaluate the performance of data science agents across realistic tasks. The paper highlights the limitations of current data science benchmarks, which often rely on simplified settings that do not accurately reflect real-world data science challenges. Key Contributions 🔹 Multimodal Tasks: DSBench includes not just textual inputs, but also images, tables, and large Excel files. Real-world data science tasks frequently involve multimodal inputs, which need to be understood holistically. 🔹Long Contexts and Complex Structures: Unlike the short, simple problems in other benchmarks, DSBench requires agents to handle long task descriptions and deal with multi-table and large data files. 🔹Realistic Data Tasks: The benchmark includes 466 data analysis tasks and 74 data modeling tasks collected from popular data science competitions like Kaggle and Eloquence. The tasks range from simple data cleaning and manipulation to full-fledged predictive modeling problems. 🔹End-to-End Evaluations: Instead of focusing on code generation or solving isolated steps, DSBench evaluates the entire workflow, from understanding the problem to producing a complete, actionable solution. This involves tasks such as model building, self-debugging, and the use of tools like Python and Excel. Concluding Remarks : The authors conclude that although significant progress has been made in the field of data science agents, there is still a considerable gap between the capabilities of current LLMs and the expertise required to solve real-world data science problems. DSBench serves as a foundation for further research aimed at developing more intelligent, autonomous, and capable data science agents. The paper encourages future work on agent-based approaches and integration with practical tools to push the boundaries of what data science agents can achieve. #GenAI #AI #LLM #datascience #machinelearning #Agents Reference : https://lnkd.in/dyzwpD8F

"Unlock the power of data! 🚀 Dive into the world of data science and uncover limitless opportunities. Check out my latest blog on how you can get started today. #DataScience #BigData #TechSkills #AI" https://lnkd.in/gPgdumJ5

📖 Read my latest article about data quality for time-series - a step by step tutorial on how you can prepare your data for forecasting! I'm passionate about time-series analysis and forecasting model development, so writing an article on this subject is always something that get me excited! After experimenting with Temporian (by Tryolabs), I became truly stoked about its potential when combined with ydata-profiling. For that reason I wanted to share this approach with the community, showcasing its efficiency and optimal processes for preparing time-series data. Ian Spektor accepted my challenge and help me to put up together the latest tutorial that I've share in Towards AI. 📘 https://lnkd.in/eUbHkgZ2 If you're curious about time-series analysis or want to see how data science can make a real-world impact, give it a read! #DataScience #TimeSeriesAnalysis #EnvironmentalAwareness #MachineLearning #data #dataquality

Great addition for Citizen Data Scientists and Applications Developers alike. Learn how #AI Quick Actions, a new feature in #OCI Data Science, makes it easy to deploy, fine-tune, and evaluate foundation models. https://lnkd.in/e9fBuh34

Just published a new blog: "Common Mistakes Beginners Make in Data Science and How to Avoid Them"! 🧑💻 Are you stepping into the world of data science? Before you dive headfirst into the algorithms, let’s chat about some rookie mistakes that many of us make—and how to avoid them. From the importance of data cleaning to the need for solid validation, I share real insights (with a sprinkle of humor) to help you on your journey. 👉 Check it out here: https://lnkd.in/dgx2yaPw Let’s learn from our mistakes and grow together in this exciting field! 💡 #DataScience #CareerAdvice #LearningJourney #DataAnalytics #DataCleaning #MachineLearning #DataPreparation #BigData #DataVisualization #DataWrangling #DataAnalysis #AI #Python #DataDriven #DataEngineer #Statistics #TechCareers #LinkedIn

An article I co-authored with Fabiana Clemente just got published in Towards AI! 💪🏼 It was a pleasure to collaborate with people as skilled and passionate about data-centric AI as Fabiana and her team at YData are. And it was also a pleasure to see how smoothly #YDataProfiling and #Temporian integrate 🤝🏼 to enable quick, easy, and thorough analysis and preparation of a pollution time series dataset! Check out the article at the link below 👇🏼