Icebreaker

🛠️⚙️EKOS ENGINEERING UPDATE #6 BUILDING THE FUTURE OF ENTERPRISE KNOWLEDGE The next generation of enterprise AI won’t be defined by models alone. It will be defined by how intelligently systems can understand, connect, exchange, and verify the knowledge behind an entire organization. That’s the future we’re building toward with EKOS. We’ve taken another important engineering step by introducing interoperability with Apache Ossie, an open format for semantic data models. Here’s what’s taking shape: 🔹 KNOWLEDGE WITHOUT BOUNDARIES EKOS can now import and export semantic models containing datasets, fields, relationships, and business context — moving enterprise knowledge toward a more open and connected ecosystem. 🔹 TRUST BUILT INTO THE ARCHITECTURE When imported models conflict with actual database schemas, EKOS identifies those inconsistencies instead of silently accepting them. 🔹 HUMAN GOVERNANCE BY DESIGN Semantic changes can be reviewed, approved, edited, or rejected, while decisions remain traceable and auditable. 🔹 VALIDATED BEYOND THE CODE We tested the end-to-end workflow using a TPC-DS model and the official Ossie validator. Two deliberately introduced schema conflicts were successfully detected during validation. 🔹 ENGINEERING FOR WHAT COMES NEXT End-to-end testing, automated validation, and CI checks are strengthening the foundation as this capability evolves. BUT THE BIGGER VISION GOES BEYOND INTEROPERABILITY. We believe the next generation of AI agents will need more than access to enterprise data. They will need a connected, structured, evidence-backed understanding of how an organization actually works — its systems, relationships, processes, and history. That requires knowledge that can move across systems without losing its meaning, evidence, or trustworthiness. Open standards make knowledge portable. Evidence makes it verifiable. Governance makes it trustworthy. EKOS is building toward a new foundation for enterprise AI — one where intelligence is grounded in knowledge that organizations can inspect, understand, and trust. We’re not just connecting data. We’re building the knowledge infrastructure for what comes next. And we’re only getting started. 📌🧷 @TheASF @IBM @Dell @UberEng @Microsoft @googlecloud @intel @databricks @awscloud @WalrusProtocol @OpenAI @AnthropicAI

@UberEng work on Apache Spark demonstrates what performance engineering can achieve at massive scale: up to 3.4x performance improvements, 79% lower runtime, and 98% lower resource costs in selected workloads. But there is a deeper engineering lesson here: faster execution does not eliminate every bottleneck. Understanding the systems behind the data is equally important.⚙️📑 In this thread, we examine Uber’s engineering approach, the challenges that remain beyond raw processing speed, and how EKOS addresses a different but complementary part of the problem. We’ll explore EKOS’s Knowledge Compiler architecture, which transforms source code, SQL, database structures, ETL pipelines, Git history, and documentation into structured, evidence-backed engineering knowledge.🛠️ We’ll also introduce its two product directions: • EKOS Lakehouse — focused on data engineering, pipelines, migration, and data reconciliation.🗄️ • EKOS System Intelligence — focused on understanding complex software environments and the relationships between code, databases, and documentation.🗄️ These are complementary engineering layers, not competing solutions. This is a conceptual comparison, not an announcement of any partnership or integration with Uber. The central idea: modern data infrastructure needs both efficient execution and a deep, verifiable understanding of the systems behind it. 📌 🧵 📎 @UberEng @Uber @ekosproject @OMID_0909 @dkhos

🚀 EKOS 1.3.0 IS HERE. FROM ENTERPRISE KNOWLEDGE TO INTELLIGENT SQL GENERATION. A major milestone for EKOS. With version 1.3.0, we’re taking another step toward making enterprise knowledge directly useful to AI agents. 🧠 WHAT’S NEW? 🔹 Text-to-SQL Agent Turn natural-language questions into SQL using compiled knowledge of the underlying database. 🔹 SQL Verification Check generated SQL against real database structures, relationships, columns, and dialect-specific rules. 🔹 Multi-Database Support Expanded SQL generation and validation across PostgreSQL, Oracle, ClickHouse, SQLite, MySQL, SQL Server, Snowflake, Databricks, and Trino. 🔹 MCP Integration Bring EKOS SQL-checking capabilities directly into AI agent workflows. 🔹 Benchmark Infrastructure More robust evaluation, reproducibility, checkpointing, and support for large-scale Text-to-SQL testing. ⚡ WHY DOES THIS MATTER? AI agents don’t just need access to enterprise data. They need to understand the systems behind it — their structures, relationships, constraints, and context. That’s what EKOS is built to provide. We believe the next generation of enterprise AI will need more than powerful models. It will need reliable knowledge infrastructure. 🛠️ EKOS 1.3.0 COMPILE KNOWLEDGE. VERIFY SQL. UNLOCK INTELLIGENCE. And we’re just getting started. 🔗 🧵

🧠 Does enterprise AI really need bigger models — or better context? One of the biggest challenges in enterprise AI isn’t just model intelligence. It’s understanding the environment in which AI operates. Enterprise knowledge is scattered across databases, source code, pipelines, documentation, Git history, business rules, and legacy systems. Giving an AI agent access to all this information doesn’t automatically mean it understands how everything connects. This is where structured knowledge becomes essential. In our latest article, we explore two independent approaches addressing this architectural challenge: 🔹 NVIDIA Auto Ontology focuses on building semantic, graph-based reasoning layers that help AI agents understand business concepts, database relationships, and enterprise data. 🔹 EKOS approaches the problem from a broader perspective: compiling knowledge across enterprise systems, resolving identities and relationships, preserving historical changes, and connecting information to verifiable evidence. These are different projects with different scopes — not a partnership or official collaboration. But they highlight an important direction for the industry: Enterprise AI needs more than powerful models. It needs reliable, structured, and reusable knowledge.⚙️🛠️ We also examine EKOS’s BIRD Mini-Dev benchmark evaluation, where the clean hold-out configuration achieved 63.9% execution accuracy compared with 53.3% for the raw DDL baseline.📑 The model remained the same. The context and surrounding workflow changed. The future isn’t just about building bigger models. It’s about giving intelligence something reliable to understand. 📖 Read the full article by Omid Ahmadi: @nvidia @NVIDIAAI @NVIDIAAIDev @JensenHuang @OMID_0909 #EnterpriseAI #NVIDIA #EKOS #AIInfrastructure #KnowledgeEngineerin

🔐 MAJOR SECURITY & CODE QUALITY REVIEW WITH SONARQUBE CLOUD! 🚀 EKOS HAS COMPLETED A MAJOR ENGINEERING HARDENING PASS, USING SONARQUBE CLOUD TO IDENTIFY AND ADDRESS CODE QUALITY AND SECURITY FINDINGS ACROSS THE PROJECT. This is more than another development milestone. It’s about strengthening the foundations of enterprise AI infrastructure. 🛡️ WHAT DID WE WORK ON? Our latest engineering effort covered findings from SonarCloud, including 96 previously open issues and 57 newly identified findings. 🔹 Security & Supply-Chain Hardening 🔹 GitHub Actions pinned to specific commit hashes. 🔹 Enforced HTTPS for downloads and strengthened S3 transport policies. 🔹 Reduced unnecessary Kubernetes Service Account permissions. 🔹 Hardened Azure disk access configurations. ⚙️ INFRASTRUCTURE & RELIABILITY 🔹 Improved Bash script safety and argument handling. 🔹 Strengthened Docker build processes. 🔹 Refined asynchronous I/O operations in the benchmarking workflow. 🧠 CODE QUALITY & MAINTAINABILITY 🔹 Reduced cognitive complexity in critical Rust components. 🔹 Refactored important Build and Commit operations. 🔹 Improved code readability and maintainability. 💡 WHY DOES THIS MATTER? EKOS is being built for complex enterprise environments — databases, application code, data pipelines, cloud infrastructure, and AI agents. That requires more than adding new features. It requires continuously strengthening the security, reliability, and engineering quality behind every layer. 🔥 WE DON’T JUST BUILD NEW CAPABILITIES. WE STRENGTHEN THE FOUNDATIONS THEY RUN ON. 🚀 EKOS | PHASE 3 DEVELOPMENT UPDATE #5 🔗 🧵 @SonarSource @ekosproject @googlecloud @Azure @intel @IBM