dataaaaa!

PuppyGraph is a zero-copy graph query engine that enables Cypher and Gremlin traversals directly on data lakehouse formats like Iceberg and Delta Lake without ETL. Its edge-sharded, MPP architecture handles super nodes and multi-hop queries at scale.

AI is forcing data tooling companies into existential pivots, with many rebuilding as AI-native to survive. While senior practitioners gain superpowers, juniors face a brutal market. AI is scrambling the playbook for vendors, practitioners, and leaders alike. The industrial-age org chart is dying.

FastBCP optimizes data exports from relational databases to cloud storage by using intra-table parallelism. It segments large tables into chunks processed simultaneously by multiple CPU cores. A treatment initially sized at 15 hours was reduced to 8 minutes with parallelization.

AI agents are disrupting data consultancy by replacing traditional staffing with agentic development. Chris Tabb explains how this shift impacts billing and prompt management. They're delivering faster, cheaper, and better than ever by leveraging internal tribal knowledge over manual labor.

Luiza Labs manages 2,000 Airflow pipelines and four petabytes of data using a YAML wrapper that enables non-engineers to deploy DAGs. They utilize custom Spark operators to optimize Kubernetes resource allocation and costs. Metadata is the most valuable asset in the AI era.

Lakebase is built on the architectural principle of separating compute and storage within OLTP databases. Originally developed at Neon, this engine enables high-performance data handling. Shamgunov describes it as separating compute and storage in an OLTP database to optimize modern workloads.

Eric Weber argues AI FOMO is manufactured by miscalibrating the industry against a tiny inner circle. While code output explodes, the real bottleneck for data engineers is judgment over throughput. He notes judgment, not throughput, is the real bottleneck right now as products remain boring.

ThoughtSpot is evolving business intelligence through an Agentic Analytics Platform that enables natural language data analysis. Key to deployment is a robust context layer to ensure precision. This shifts the data analyst's role toward becoming an enabler of automated agentic systems.

France faces a critical infrastructure challenge regarding the number of datacenters required to support growing AI and data processing needs. Discussions highlight the technical necessity of expanding local capacity to maintain digital sovereignty and handle increasing computational loads.

Suresh Srinivas explores the evolution from big data to metadata management. He argues that metadata is essential for scaling data operations and improving discovery. Suresh Srinivas joins us to discuss his journey from big data to metadata, highlighting how structured context drives value.

Octave Deschamps demonstrates how to sustain a high-level Power BI freelance career while sailing the Atlantic. He relies on a strategy of short, 100 percent remote missions and regular clients to maintain stability. He decides to buy a sailboat and cross the Atlantic for 10 months.

AI analytics is transforming the data landscape, shifting the focus from manual reporting to reliability and internal tool building. Tools like Omni, Hex, and Claude Code are driving this acceleration. The future of analytics may move toward a dashboardless environment.

Veronika Durgin argues that AI shifts data engineering from production to translation and judgment. Analytics engineering is becoming the central catch-all role. Conceptual data modeling as the unglamorous prerequisite to AI is essential for defining business logic within semantic layers.

Apache Iceberg V3 introduces vendor-neutral definitions for semi-structured data and efficient delete representations to optimize batch analytics. The community is now developing the V4 specification to support streaming and AI workloads through one-file commits and improved columnar statistics.

Iceberg v4 introduces adaptive metadata trees to reduce write amplification and commit latency for streaming workloads. By shifting manifest storage from Avro to Parquet, the architecture enables columnar stats projection and one-file commits.

Cargill is scaling Airflow across 60+ teams via its Minerva platform to replace legacy ETL. Key strategies include using native SLA sensors for batch ingestion and enforcing single-purpose tasks. Every task in a DAG should have a single, documented purpose to simplify troubleshooting.

OntoBricks is an open-source solution that integrates with Databricks to transition from technical governance to semantic knowledge organization. It uses Knowledge Graphs to create an organizational digital twin.

Amit Sharma proposes collapsing the modern data stack to streamline AI integration. By reducing architectural layers, engineers can better support real-time AI applications. He focuses on a vision for collapsing the modern data stack to better support AI applications through simplified data access.

Scaling agentic AI training requires overcoming barriers in ROI, security, and skills. Success relies on a hybrid model of e-learning and mentoring to ensure data privacy and technical sovereignty. The four barriers to AI adoption in companies: skills, ROI, security, and sovereignty.

Teradata has evolved from an MPP legacy into a knowledge platform that integrates autonomous AI with unstructured data processing. Its hybrid architecture ensures consistent deployment across cloud and on-premise environments. The major strength of their solution lies in its hybrid architecture.

Dr. Fern Halper discusses her latest research on the evolving landscape of data analytics and AI. The conversation focuses on how organizations integrate advanced analytics into their core infrastructure. Halper highlights that data makes the world go round by driving modern decision-making.

DuckDB is evolving into a client-server model with the introduction of Quack, expanding its utility beyond local processing. Additionally, Mistral CEO Arthur Mensch addressed French deputies regarding digital sovereignty.

Exotec is overhauling its data architecture to enable self-service and decentralization. Head of Data Gregory Vanuynsberghe focuses on building stakeholder trust and deploying specialized analysts by domain. Key goals include a stack redesign to support a dashboardless future and GenAI.

Data engineers must move beyond vague ethics to identify specific technical disasters like cascading failures in agentic AI. Success is measured by avoiding emergent risks through cross-functional governance.

Data infrastructure is merging with AI as companies pivot toward becoming AI-native layers for context, orchestration, and compute. Benchmarks like ADE-bench show models excel at pipeline building, yet current tests often fail to capture real-world statefulness.