Davide Mauri’s Post

Exciting updates from #MicrosoftBuild this week!

How does MongoDB make building AI apps easier and faster? (Hint: It starts with unifying operational and vector data) Learn more here: https://lnkd.in/gGP_AJwE #LoveYourDevelopers #mongodb #nosql #sql #genai #ai #database #developer #llm #RAG #architect #vector #search #mongodbatlas

Did you know that 80% of enterprise developers are either testing or planning to test #GenAI within their workflows and routines? 💡 This latest IDC report analyses MongoDB's efforts to simplify the integration of large language models (#LLMs) with organizational data, empowering developers to create #GenAI-based apps that drive business value. Read more. https://lnkd.in/g8mZqJXg #mongodb #nosql #sql #genai #ai #database #developer #llm #RAG #architect #vector #search #mongodbatlas

Build #AI apps faster and easier with #SQL #database in #Fabric – now Public Preview! https://buff.ly/3AOZhxd #Microsoft #SQLServer #AzureSQL @SQLServer #MadeiraData

🚀 Working with Delta Tables in Microsoft Fabric 🚀 Delta tables in Microsoft Fabric bring the power of relational database features to the flexibility of data lakes. With Delta tables, you can efficiently manage both structured and unstructured data using Apache Spark. Here's a quick guide on how to create and manage these tables: Creating Delta Tables: From a DataFrame: You can easily save a DataFrame as a Delta table using Spark. For example, the following PySpark code loads a CSV file and saves it as a Delta table: python Copy code df = spark.read.load('Files/mydata.csv', format='csv', header=True) df.write.format("delta").saveAsTable("mytable") The data is stored in Parquet files and managed by Delta Lake with transactional logging (_delta_log). Managed vs. External Tables: Managed Tables: The table definition and data files are both managed by the Spark runtime. Deleting the table also deletes the files. External Tables: You can store the data in custom locations by specifying a path, allowing you to control file storage independently from the table definition. Benefits: ACID Transactions: Delta tables provide atomicity, consistency, isolation, and durability for reliable data management. Schema Enforcement: Structured data follows defined schemas to ensure consistency. Version Control & Time Travel: Track changes in data with transaction logs and revert to earlier versions if needed. Batch & Streaming Support: Handle real-time data with ease, supporting both batch and streaming pipelines. Delta tables combine the best of both worlds: the scalability of data lakes and the structure of databases, making Microsoft Fabric an ideal platform for modern data engineering. #DataEngineering #DeltaLake #MicrosoftFabric #ApacheSpark #BigData #CloudComputing #DataManagement

In Databricks SQL and Unity Catalog, you can create and manage AI functions with AI_QUERY and SQL Functions to make accessing models and standard prompting templates a breeze in SQL, but did you also know you can create embeddings and query them directly on your tables in SQL with VECTOR_SEARCH? Check this blog by Reilly Wild-Williams that walks through how to use all these tools to create AI-embedded SQL Dashboards and pipelines! https://lnkd.in/gx4GqnXr

From modernizing apps to building new AI-powered experiences, developers are key. Learn how we put developers first: https://lnkd.in/gGP_AJwE #LoveYourDevelopers MongoDB #mongodb #nosql #sql #genai #ai #database #developer #llm #RAG #architect #vector #search #mongodbatlas

SQL Agent vs RAG: Which One Wins for Tabular Data?   Just wrapped up an exciting project where I compared Retrieval-Augmented Generation (RAG) against SQL Agent for working with tabular datasets – and the results were surprising! For this project, I used Azure OpenAI API for embedding and chatbot integration, LangChain SQL Agent for SQL handling, and SQLite as the relational database. I also incorporated ChromaDB as the vector database for RAG-related tasks and built the web interface using Streamlit to make everything accessible and user-friendly. While RAG has its strengths with unstructured and hybrid data, SQL Agent clearly outshined when it came to structured, tabular datasets.   Here’s what stood out: Performance: RAG had a faster response time, but SQL Agent provided perfect answers most of the time. Accuracy: Direct SQL queries maintained high precision, especially with relational data. Scalability: SQL-based solutions scaled better with larger tables, reducing the operational complexity associated with embedding-heavy RAG setups. RAG is still a powerful tool for dynamic and contextual queries, but SQL Agent proved to be the clear winner for this use case.   Key takeaway: Choosing the right tool is all about understanding your data and selecting solutions that align with the task at hand. I’d love to hear suggestions for RAG applications in structured data – I'm curious to explore them! For my next use case, I plan to use RAG with GraphDB, combine RAG and SQL Agent to compare their performances.   github: https://lnkd.in/gfuPF3VB #RAG #SQLAgent #AI #MachineLearning #TabularData #AzureOpenAI #ChromaDB #LangChain #DataScience #AIEngineer

🚀 Vector Databases vs. Traditional Databases: A Quick Comparison 🚀 As data continues to grow in both volume and complexity, choosing the right type of database for your use case is crucial. Below is a comparison between Vector Databases and Traditional Databases to help you understand their differences and which one might be best for your needs. 💡 Key Takeaway: If you’re working with AI, machine learning models, or similarity-based queries, Vector Databases offer superior performance for tasks like search and recommendation. On the other hand, for traditional transactional use cases, Traditional Databases like MySQL or PostgreSQL continue to shine with strong consistency, relational integrity, and scalability. Which database solution do you prefer for your projects? Drop your thoughts in the comments below! 👇 #Database #VectorDatabases #MachineLearning #AI #DataScience #Technology #BigData #SQL #PostgreSQL #MySQL #DataManagement

Improving Text2SQL Performance with Ease on Databricks https://zurl.co/9578 << The Spider benchmark is a widely recognized standard for evaluating the performance of text-to-SQL systems. It challenges #LLMs to translate natural language queries into precise #SQL statements