Get a 30% prepaid usage bonus on Apify this Black Friday.

Scrape data for LLMs, machine learning, competitive intelligence, product mapping, or any AI use cases.

Use ready-made scrapers or build your own.

The Boost plan ends December 5 - grab it before it’s gone!

Sponsored

🗞️Welcome to DataPro #121 – Your Weekly DS & ML Highlights! 🌟

Stay on top of the ever-evolving world of Data Science, AI, and ML! This week, we’ve curated the hottest resources, tools, and breakthroughs to empower your projects and sharpen your skills. Let’s explore!

🔍 Top Picks: Must-Know Insights for Data Pros

◘ Cortex for Local LLMs: Simplify running local language models.

◘ AnythingLLM: Your all-in-one LLM app.

◘ Smarter Maps with GPT-4o: Explore fine-tuning for advanced geospatial tools.

◘ AI for Good: Tackling real-world challenges with Yasuyuki Matsushita at Microsoft Research Asia.

◘ BiomedParse: Revolutionize biomedical image analysis with this foundation model.

◘ Orca-AgentInstruct: Harness synthetic data through agentic flows.

◘ GraphRAG: Boost global search with dynamic community selection.

🚀 Next-Level Tech Trends

◘ Google Cloud’s Translation AI Updates: Breaking boundaries in translation technology.

◘ Caravan MultiMet by Google AI: Exploring multi-model alignment.

◘ Infinite-Length Video Generation: Dive into "Meet The Matrix."

◘ FluidML: Smarter runtime management for ML inference.

◘ AWS Multi-Agent Orchestrator: Seamlessly manage AI agents.

◘ DeepSeek’s Reasoning Engine: Unveiling DeepSeek-R1-Lite-Preview.

◘ Pixtral Large: Mistral AI’s 124B multimodal innovation.

◘ XiYan-SQL by Alibaba Research: The ultimate Text-to-SQL framework.

◘ Lingma SWE-GPT: Open-source solutions for software development challenges.

🛠️ ML Tools & Tactics

◘ AI-Powered Prompt Writing: Save time with smarter designs.

◘ NER with Hugging Face: A simple guide to Named Entity Recognition.

◘ 10 Python Libraries: Essential tools for data analysts.

◘ ETL Pipelines: Develop robust workflows for data projects.

◘ Advanced SQL Techniques: Master data manipulation like a pro.

◘ Python + DuckDB: Speed up your data analysis.

◘ Google Cloud Data Security: A guide to building a secure platform.

📊 In Action: Real-World ML Wins

◘ Why AI Strategies Fail: Common pitfalls and how to avoid them.

◘ Data-Driven Customer Systems: Build better management frameworks.

◘ Research Assistants for Docs: Automate document creation with AI.

◘ Feature Engineering in Healthcare: Transform insights with smart techniques.

◘ 3D Imaging with Nvidia LLaMa-Mesh: Bring your visuals to life.

◘ Multimodal Models: LLMs that see and hear.

◘ Understanding Data Labeling: A hands-on guide.

◘ Cost Savings with Ray on Amazon EKS: How Vannevar Labs cut ML costs by 45%.

🌍 Industry Buzz & Discoveries

◘ Optimizing Transformers: Make attention layers work harder.

◘ Neural Network Quantization: Tips to streamline your models.

◘ Email Automation with Amazon Bedrock: Smarter Q&A workflows.

◘ Integrated Text & Image Classification: Next-gen data analysis.

◘ NetworkX in Python: Master graphs and networks with ease.

◘ Fixing Cross-Validation Visuals: Avoid common pitfalls in data visualization.

Stay tuned and stay inspired – there’s always something new to discover in the ever-evolving world of Data Science and Machine Learning!

Take our weekly survey and get a free PDF copy of our best-selling book,"Interactive Data Visualization with Python - Second Edition."We appreciate your input and hope you enjoy the book!

Share Your Insights and Shine! 🌟💬

Cheers,

Merlyn Shelley,

Editor-in-Chief, Packt.

➽ RAG-Driven Generative AI: This new title, RAG-Driven Generative AI, is perfect for engineers and database developers looking to build AI systems that give accurate, reliable answers by connecting responses to their source documents. It helps you reduce hallucinations, balance cost and performance, and improve accuracy using real-time feedback and tools like Pinecone and Deep Lake. By the end, you’ll know how to design AI that makes smart decisions based on real-world data—perfect for scaling projects and staying competitive! Start your free trial for access, renewing at $19.99/month.

➽ Building Production-Grade Web Applications with Supabase: This new book is all about helping you master Supabase and Next.js to build scalable, secure web apps. It’s perfect for solving tech challenges like real-time data handling, file storage, and enhancing app security. You'll even learn how to automate tasks and work with multi-tenant systems, making your projects more efficient. By the end, you'll be a Supabase pro! Start your free trial for access, renewing at $19.99/month.

➽ Python Data Cleaning and Preparation Best Practices: This new book is a great guide for improving data quality and handling. It helps solve common tech issues like messy, incomplete data and missing out on insights from unstructured data. You’ll learn how to clean, validate, and transform both structured and unstructured data—think text, images, and audio—making your data pipelines reliable and your results more meaningful. Perfect for sharpening your data skills! Start your free trial for access, renewing at $19.99/month.

⫸ Run Local LLMs with Cortex: This blog introduces Cortex, a tool that allows you to run and customize local LLMs easily on your machine. It guides you through installation, model selection, and usage, making AI accessible even with standard hardware.

⫸ AnythingLLM: The LLM Application You’ve Been Waiting For. This blog introduces AnythingLLM, an open-source platform that helps you build private ChatGPT-like agents. It offers advanced capabilities, privacy, and flexibility, with step-by-step instructions on getting started for various use cases.

⫸ Building smarter maps with GPT-4o vision fine-tuning: This blog highlights Grab's innovative use of GPT-4o vision fine-tuning to improve its mapping service, GrabMaps. By enhancing localization and automation in mapmaking, Grab reduces costs, increases accuracy, and boosts data trust for Southeast Asia's dynamic landscape.

⫸ Tackling societal challenges with AI at Microsoft Research Asia - Tokyo. This blog celebrates the opening of Microsoft Research’s new Tokyo lab, focusing on embodied AI, societal challenges, and industry innovation. Led by Yasuyuki Matsushita, it aims to drive local and global AI advancements through collaboration and talent development.

⫸ BiomedParse: A foundation model for smarter, all-in-one biomedical image analysis. This blog introduces BiomedParse, an advanced framework for holistic biomedical image analysis. It unifies object recognition, detection, and segmentation into a single model, offering faster, more accurate insights by using natural-language prompts for medical image analysis.

⫸ Orca-AgentInstruct: Agentic flows can be effective synthetic-data generators. This blog introduces Orca-AgentInstruct, an agentic framework for generating diverse, high-quality synthetic data to fine-tune language models. By leveraging agentic flows, it enables scalable, autonomous data generation, leading to substantial performance improvements across multiple benchmarks.

⫸ GraphRAG: Improving global search via dynamic community selection. This blog introduces GraphRAG, an advanced method for handling "global" queries using a dynamic global search. It efficiently utilizes a hierarchical knowledge graph, reducing costs by pruning irrelevant community reports and improving response quality for abstract questions.

⫸ Sharing the latest updates to Google Cloud’s Translation AI: This blog introduces Google Cloud's Translation AI in Vertex AI, offering advanced tools for accurate, customizable translation. It includes two options: Translation API Basic for speed and consistency, and Translation API Advanced for tailored, high-quality translations at scale.

⫸ Google AI Research Introduces Caravan MultiMet: This blog introduces the Caravan MultiMet extension, a breakthrough in large-sample hydrology. By integrating real-time forecast and nowcast data into the Caravan dataset, it enhances hydrological model accuracy, improving forecasting, benchmarking, and water resource management.

⫸ Meet The Matrix: A New AI Approach to Infinite-Length and Real-Time Video Generation. This blog introduces The Matrix, a groundbreaking world model for generating infinite-length, real-time video simulations with high fidelity. It uses advanced diffusion techniques to enable interactive, scalable simulations across both game and real-world environments, revolutionizing video generation for gaming, training, and VR.

⫸ Meet FluidML: A Generic Runtime Memory Management and Optimization Framework for Faster, Smarter Machine Learning Inference. This blog introduces FluidML, an advanced framework designed to optimize machine learning inference on edge devices. By improving memory layout, graph segmentation, and scheduling, FluidML achieves significant reductions in latency and memory usage, enabling real-time deployment of complex models in resource-constrained environments.

⫸ AWS Releases 'Multi-Agent Orchestrator': A New AI Framework for Managing AI Agents and Handling Complex Conversations. This blog introduces AWS's Multi-Agent Orchestrator, a framework designed to manage multiple AI agents. It intelligently routes queries, maintains context, and supports flexible deployment across various environments, enhancing the scalability and coherence of conversational AI systems.

⫸ DeepSeek Introduces DeepSeek-R1-Lite-Preview with Complete Reasoning Outputs Matching OpenAI o1. This blog introduces DeepSeek-R1-Lite-Preview, a model designed to enhance transparency in AI reasoning. By incorporating Chain-of-Thought capabilities, it provides step-by-step explanations for complex tasks, improving trust and understanding in AI-driven problem-solving.

⫸ Mistral AI Releases Pixtral Large: A 124B Open-Weights Multimodal Model Built on Top of Mistral Large 2. This blog introduces Pixtral Large, a 124 billion-parameter multimodal AI model by Mistral AI. Built on Mistral Large 2, it integrates text, images, and other data types, offering open weights for customizable research and application development.

⫸ Alibaba Research Introduces XiYan-SQL: A Multi-Generator Ensemble AI Framework for Text-to-SQL. This blog introduces XiYan-SQL, an innovative NL2SQL framework that enhances query generation through multi-generator ensemble strategies and advanced schema representation. With superior performance across multiple benchmarks, it balances accuracy, adaptability, and diversity for complex database interactions.

⫸ Lingma SWE-GPT: Pioneering AI-Assisted Solutions for Software Development Challenges with Innovative Open-Source Models. This blog introduces Lingma SWE-GPT, an open-source LLM series designed for software engineering tasks. With improved fault localization, patch generation, and iterative reasoning, it bridges performance gaps between open and closed-source models while remaining cost-effective and scalable.

⫸ Save time on prompt design with AI-powered prompt writing: This blog introduces new features in Vertex AI to simplify prompt engineering: "Generate prompt" for quickly creating effective prompts based on objectives, and "Refine prompt" for improving them with AI-driven suggestions, streamlining the workflow and enhancing prompt quality.

⫸ How to Implement Named Entity Recognition with Hugging Face Transformers? This blog demonstrates how to perform Named Entity Recognition (NER) using Hugging Face’s Transformers library. By using a pre-trained BERT model fine-tuned for NER tasks, the tutorial walks through tokenization, entity identification, and results interpretation, helping developers extract valuable insights from text.

⫸ 10 Python Libraries Every Data Analyst Should Know: This blog highlights essential Python libraries for data analysts. It covers tools for data retrieval (Requests, Beautiful Soup), manipulation (NumPy, Pandas, Polars), statistical analysis (Statsmodels, SciPy), and visualization (Seaborn), along with database interaction (SQLAlchemy), all aimed at simplifying and enhancing the data analysis workflow.

⫸ Developing Robust ETL Pipelines for Data Science Projects: This blog introduces the process of building an ETL pipeline for data science projects. It covers the steps of Extracting, Transforming, and Loading data, using Python libraries like Pandas and SQLite to automate the data cleaning and storage process for efficient analysis.

⫸ 7 Advanced SQL Techniques for Data Manipulation in Data Science: This blog highlights seven advanced SQL techniques for data manipulation in data science. These techniques include subqueries, correlated subqueries, Common Table Expressions (CTEs), and recursive queries, all of which help streamline complex queries, restructure data, and handle hierarchical data efficiently.

⫸ A Guide to Data Analysis in Python with DuckDB: This blog introduces DuckDB, an in-process OLAP database for analyzing data in Python. It demonstrates how to set up the environment, install DuckDB, and query data from CSV files using SQL, making data analysis with pandas and other data sources more efficient.

⫸ Learn how to build a secure data platform with Google Cloud ebook: This blog explores how Google Cloud's data security tools can protect your business data while fostering innovation. It covers encryption, access controls, compliance, and monitoring to help safeguard your data in today’s complex security landscape.

⫸ The Root Cause of Why Organizations Fail With Data & AI: This article explains why many companies struggle to monetize their data and how the lack of a clear business strategy is the root cause. It emphasizes the importance of aligning business strategies with data initiatives for success.

⫸ How to Build a Data-Driven Customer Management System: This article explores how customer base management (CBM) systems help businesses optimize pricing, predict churn, and enhance decision-making. It covers foundational components like ELT, churn modeling, and dashboards, and examines how advanced features can provide a strategic edge.

⫸ Building a Research Assistant That Can Write to Google Docs: This article, part two of a series, explains how to connect a research agent to Google Docs using LangGraph and Tavily. It covers setting up Google Drive and Docs APIs, creating folders, and uploading documents programmatically.

⫸ Feature Engineering Techniques for Healthcare Data Analysis: This article continues a feature engineering project focused on healthcare data, specifically on handling patient diagnosis data to uncover hidden insights. It highlights the importance of domain knowledge in transforming raw data, using techniques like comorbidity analysis to create meaningful features for better predictions and outcomes.

⫸ Generate 3D Images with Nvidia’s LLaMa-Mesh: This article explores NVIDIA's LLaMA-Mesh, a model that generates 3D mesh objects from natural language descriptions. It highlights how vertex quantization and OBJ format enable seamless 3D object creation and understanding, with applications across various industries.

⫸ Multimodal Models — LLMs That Can See and Hear: This article introduces multimodal AI, focusing on models that combine text and image processing. It explores using LLaMA 3.2 Vision for image-to-text tasks like visual question answering, demonstrating the power of LLMs in handling multiple modalities.

⫸ Understanding Data Labeling (Guide): This article explains the importance of data labeling in machine learning, discussing its role in supervised learning, types of labeling (e.g., image classification, sentiment analysis), and various approaches, including human-in-the-loop and automated methods.

⫸ How Vannevar Labs cut ML inference costs by 45% using Ray on Amazon EKS? This post details how Vannevar Labs optimized its ML inference workloads using Ray, Karpenter, and Amazon EKS, achieving a 45% reduction in costs. They employed Ray Serve for efficient inference, used Karpenter for optimized instance selection, and leveraged fractional GPUs for improved resource utilization.

⫸ Increasing Transformer Model Efficiency Through Attention Layer Optimization: This article explores optimization techniques for attention layers in Transformer models using PyTorch. It covers various methods like PyTorch SDPA, FlashAttention, and third-party solutions such as Transformer Engine to enhance computational efficiency and reduce resource consumption, offering insights into real-world performance improvements.

⫸ Quantizing Neural Network Models: This post discusses techniques for quantizing AI models to reduce their size and computational cost while maintaining accuracy. It focuses on two methods: Post-Training Quantization (PTQ) and Quantization Aware Training (QAT), highlighting their advantages, challenges, and use cases.

⫸ Automate Q&A email responses with Amazon Bedrock Knowledge Bases: This post discusses automating email responses using generative AI, combining Retrieval Augmented Generation (RAG) and Amazon Bedrock Knowledge Bases. It outlines a solution that improves HR operations by automating email replies with accurate, contextually relevant information from company knowledge bases.

⫸ Integrating Text and Images for Smarter Data Classification: This post provides a technical guide on building a multimodal AI pipeline for classifying mixed text and image data. Using Gemini 1.5 and LangChain, the tutorial covers setting up the system for image-text classification, including key steps like defining output schemas, encoding image data, and handling structured outputs for accurate classification.

⫸ Navigating Networks with NetworkX: A Short Guide to Graphs in Python. This post introduces NetworkX, a powerful library for building, analyzing, and visualizing graphs. It explains how to create graphs, add nodes and edges with attributes, and visualize them using Matplotlib. The post also demonstrates these concepts with examples, including the famous Zachary’s Karate Club network.

⫸ Why Most Cross-Validation Visualizations Are Wrong (And How to Fix Them)? This post explores how current cross-validation diagrams often confuse learners and suggests a better approach. It discusses how traditional visualizations rely too much on color and movement, which mislead understanding, and offers a simpler, more intuitive design for explaining cross-validation processes.

We’ve got more great things coming your way—see you soon!