dataaaaa!

This episode examines Nvidia's acquisition of open-source model repository Hugging Face. Miriah Peterson and Matt Sharp discuss the technical implications for model hosting, deployment pipelines, and operationalizing LLMs in production environments.

dbt brings software engineering practices to data transformation by organizing SQL select statements into a DAG. It persists cleaned, modeled tables in the warehouse, while its semantic layer provides consistent metric definitions required for downstream BI tools and AI agents.

Data and AI transformation consultant Edouard Stevenini operates as Chief of Staff for a major enterprise CDO. The role oversees high-level PMO execution, decision support, and strategic management for a 400-person data organization while navigating enterprise AI adoption trends.

Colin Zima outlines the future of BI, predicting that AI will write more data models than people while semantics become increasingly vital. Data teams will continue to play a crucial role as architectures shift beyond simple SQL toward robust semantic layers.

Data modeling is shifting toward automation as AI begins writing more models than engineers. Colin Zima predicts robust semantic layers and data teams remain essential to manage business logic and SQL complexity as dashboards evolve.

AI reasoning systems are advancing via formally verified proofs using the Lean language. Tools like Aristotle solve complex mathematics by verifying logic and discovering counterexamples, driving rigorous formal verification methods that improve general technical reasoning.

Technical relevance in data engineering is evolving beyond traditional operating modes as AI reshapes what hands-on means. Rather than remaining trapped in routine pipeline tasks, roles are expanding toward architectural advising, teaching, and building next-generation data systems.

Agentic AI evolution mimics RPA with master-agent architectures, but enterprise adoption lags. High token infrastructure costs and data sovereignty regulations limit production use in regulated pipelines, leaving only 10 to 15% of implementations succeeding over human data analysts.

DuckDB is evolving its analytical engine to handle modern data engineering demands. To enable community features, developers rewrote its SQL parser and built an extension mechanism. It is also introducing client-server architecture and DuckLake as an alternative table format to Apache Iceberg.

Dan Bennett discusses using Claude Code to build an open-source RDF extension for DuckDB in production C++. He stresses robust test coverage for LLMs, autonomous agents in headless VMs, and solving cross-organizational data semantics.

Scaling language-centric models fails at complex visual reasoning like spatial analysis and technical design. Elorian AI builds specialized multimodal architectures, data pipelines, and reinforcement learning systems to overcome vision bottlenecks in robotics and CAD automation.

In agentic workflows, context and skill files act as code and need software lifecycles like versioning and evals. Teams must build custom harnesses to manage shared context across data and engineering agents, cutting token costs while avoiding frontier vendor lock-in.

Astronomer introduced Otto, a specialized AI agent designed for Apache Airflow workflows. It provides essential context for DAG authoring, code reviews, version migrations, and failure investigations, driving toward self-healing data pipelines while integrating directly into existing data

Wise manages financial data by treating Airflow as a pure orchestration layer and offloading compute to EMR or SageMaker. They isolate execution using the Python Virtual Environment Operator and handle non-idempotent LLM workflows via multi-layer validation and agent-checking-agent patterns.

LinkedIn built a four-layer memory agent to persist state for AI hiring assistants. They replaced GraphRAG with tree-structured memory for faster updates, managing conversational, semantic, and episodic layers to balance retrieval freshness and latency.

Paul Domanski demonstrates orchestrating over 50 concurrent AI agents using Tmux and Demox workspaces. The setup balances workloads across Claude and Codex accounts while managing API limits, context windows, and token efficiency for rapid enterprise software prototyping and deployment.

Engineering teams must treat skill sets with the same software lifecycle management as code to scale effectively. Guy Podjarny argues that organizations will build and own their own harnesses rather than relying entirely on rented infrastructure from frontier AI labs.

Modern Data 101 is launching The Modern Data Podcast with video discussions coming soon.

Amazon has signed an agreement to acquire DuckLabs, the developer of the DuckDB analytics database engine. This move brings DuckDB's in-process analytical query processing capabilities directly under the AWS ecosystem.

Automate BI and analytics development using AI coding agents like Claude Code. Engineering teams scale workflows by treating agents like a harness for productivity, while managing technical risks like token costs and shadow IT.

Oracle SVP Vipin Samar discusses database security architectures tailored for agentic AI. The episode focuses on securing systems directly at the database layer to enable robust, protected AI data pipelines.

Meta's Research Science teams rely on a data-driven culture centered on experimentation and metric alignment. Applied Artificial Intelligence workflows focus on scaling technical architectures and integrating robust data engineering practices to support advanced AI adoption.

This showcase explores extensions to Apache DataFusion. RDF Fusion integrates SPARQL and piecewise signal querying to join graph models with IoT time series data. Additionally, Cloudflare demonstrates how "R2 SQL is Cloudflare's serverless, distributed, analytics query engine for Iceberg catalogs."

Deploying ML models often stalls due to centralized governance rather than technical limits. AWS leaders discuss shifting from rigid control to adaptive architectures where teams make fast, reversible decisions and become data-informed to accelerate deployment pipelines.

Joe Reis discusses The Mixed Model Arts Manifesto, exploring modern data modeling concepts and practical architectures for data engineering workflows.