dataaaaa!

Use the DuckDB and Quack n8n node to convert CSV files into Parquet via in-memory SQL execution. By specifying headers and delimiters during ingestion, pipelines can transform datasets, save Parquet files locally, and query the resulting columnar data directly in downstream n8n nodes.

Data systems, software, and AI engineering are converging as AI collapses traditional discipline boundaries. As these technical roles blend together, core architectural fundamentals, semantics, specification, and human judgment become increasingly critical.

Production AI is shifting from massive frontier models toward targeted, small open-source models. Using open-source inference infrastructure, data engineers can orchestrate smart agents to build faster, cheaper, and control-driven systems as small models close the performance gap.

Silicon Data provides market intelligence for the AI compute ecosystem by tracking real-time GPU rental rates, LLM token expenditure trends, and compute forward curves. Head of Research Steve Hou notes that they aim to bring data transparency and price discovery to the physical AI compute market.

Enterprises prefer Model Context Protocol over CLIs for managing AI agents. Jeremiah Lowin explains that agentic workflows are DAGs of outcomes rather than implementations. Data teams must govern tool access securely, as "a skill file is a polite note" that agents can ignore.

Airflow orchestration decisions directly dictate Apache Spark performance and cloud costs. Misconfigurations, like improperly parallelizing tasks or ignoring underlying compute usage, can slow down jobs.

Jeremiah Lowin built FastMCP after Anthropic announced Model Context Protocol. Enterprises prefer MCP over CLIs, focusing on enterprise security plumbing before adopting context layers. Lowin notes a skill file is a polite note rather than a workflow engine for agents.

Ryan Dolley discusses integrating large language models with business intelligence in an episode recorded at DataTune 2026. The discussion focuses on insights from his upcoming book on BI and LLMs.

Accion Labs uses Apache Airflow to modernize client infrastructure, containerizing deployments on Kubernetes to avoid cloud lock-in. For a financial client, replacing cron jobs with a single Airflow DAG reduced engineering effort by 25%.

Yoon Kim of TwelveLabs explains that native video understanding requires specialized models like Marengo and Pegasus, rather than relying on LLMs that take periodic frame snapshots. To search across multiple videos — maybe hundreds of thousands of hours — it becomes a scale problem.

Rogue AI agent incidents stem from poor network security, permissions, and governance rather than unbridled intelligence. As agents become more autonomous, enterprises must prioritize access control, security, and the economics of inference and training to mitigate risks.

Building scalable AI systems requires deliberate architecture rather than just wrapping LLMs around existing workflows. Data engineering teams must carefully curate data pipelines to prevent a "semantic swamp" and deploy autonomous cloud agents driven by budget-friendly AI models.

Tech claims about the end of SQL or data warehousing rely on logical fallacies for clickbait. Core technologies evolve rather than disappear. Separating core fundamentals from changing implementations reveals where the industry is heading. "Ragebait gets the clicks and comments."

Former Data Project Lead Elise Juhel outlines transitioning to freelance data governance consulting. She details managing short contracts alongside her main role supporting Moet Hennessy, leveraging consulting agency experience and network connections to drive career growth.

Freshworks CPO Srini Raghavan discusses scaling data science and AI software architectures. System design is shifting toward modular architectures and Model Context Protocol to serve non-human users, noting "a growing share of users aren't people at all — they're AI agents calling the same systems."

Tristan Handy and Jason Ganz discuss key data ecosystem shifts, analyzing autonomous AI agents and open-weight licensing like Moonshots Kimi K3. The episode highlights "what changes for data teams as agents arrive" to evaluate how AI workflows alter modern analytics engineering practices.

DataGen covers four key data and AI trends for 2026 with Christophe Blefari. Topics include GenAI for data engineering, conversational interfaces for the Modern Data Stack, semantic layers, and why 95% of GenAI projects struggle to reach production.

As AI agents evolve, data engineering priorities are shifting toward structured platform work and context delivery. Data teams must connect agents using semantic layers like dbt MCP servers to enforce governance.

Building robust data products requires blending standard data engineering workflows with AI capabilities. Engineers focus on constructing scalable ingestion pipelines, managing unstructured data, and integrating model endpoints into production systems.

Joe Reis clarifies that his data podcast is strictly invitation-only to stop incoming pitches. He reports being flooded with low-quality PR pitch emails for AI startups, calling the automated outreach nonsensical. Unsolicited requests will be ignored or publicly berated.

Pendo CEO Todd Olson discusses using behavioral data and rage prompts to improve software. While AI coding tools speed up development, teams need rich context to build effective products. The episode covers headless and agentic design alongside the evolving roles in software engineering.

Former Data Analyst Caroline Mestre transitioned to freelance Analytics Engineering after two years of experience. Working remotely, she secures contracts with major clients by prioritizing project quality and working conditions, eventually raising her daily rate from 450 to 550 euros.

Photoroom leveraged generative AI to accelerate analytics delivery speed by 15x. Head of Data Juliette Duizabo integrated AI across the entire data pipeline, covering ingestion, transformation, the semantic layer, and analytics while altering data roles.

OmniGraph is a lakehouse-native graph storage layer designed specifically for agentic systems. Built on top of object storage, Lance, Arrow, and DataFusion, it combines graph semantics with Git-style branching and merging.

AI is transforming engineering as engineers can increasingly produce code, schemas, documentation, and infrastructure without fully understanding what they created. Software and data engineers are shifting from building pipelines to designing the semantic infrastructure machines require.