# Yukon Labs

> Enterprise AI and data platforms for governments, banks and regulated
> enterprises in Azerbaijan and across the CIS and MENA regions. Yukon Labs
> implements data governance (OvalEdge), data lakehouse and federated analytics
> (Starburst / Trino) and AI orchestration (HAVAA) — deployed on the client's own
> perimeter, including fully air-gapped environments. ISO/IEC 42001 certified.

Contact: consulting@yukon.az · Baku, Azerbaijan

## Solutions

- [Data governance solutions in Azerbaijan](https://yukon.az/solutions/data-governance): What data governance is, why regulated enterprises need it, and how Yukon Labs implements it — catalog, lineage, classification, quality and access governance.
- [Data lakehouse solutions in Azerbaijan](https://yukon.az/solutions/data-lakehouse): Modern data architecture that queries every source as one, in open table formats, without pipelines or duplicated copies of regulated data.
- [AI orchestration solutions in Azerbaijan](https://yukon.az/solutions/ai-orchestration): The platform layer that coordinates AI agents, enterprise systems and human approvals under recorded policy, so an AI deployment can be audited and defended.

## Platforms

- [OvalEdge — data governance and data catalog](https://yukon.az/solutions/data-governance/ovaledge): Active metadata across 150+ connectors, automated PII classification, column-level lineage and access workflows. Not a database: a metadata and governance layer that leaves the data in place.
- [Starburst — data lakehouse and federated query](https://yukon.az/solutions/data-lakehouse/starburst): Trino-based MPP SQL across object storage, databases and warehouses, with Apache Iceberg and Delta Lake, without moving or copying the data.
- [HAVAA — enterprise AI agent platform](https://yukon.az/solutions/ai-orchestration/havaa-ai): Yukon Labs' own agentic AI platform. Four safety modes (Chatbot, Assistant, Worker, Autonomous), human-in-the-loop approvals and a full audit trail; SaaS, Docker, Kubernetes or air-gapped.

## Reference

- [Frequently asked questions](https://yukon.az/faq): Direct answers on deployment models, data residency, pricing shape and engagement length.
- [AI policy (ISO/IEC 42001)](https://yukon.az/ai-policy): The published AI management system policy Yukon Labs is audited against.
- [Governance maturity assessment](https://yukon.az/assessment): Self-scoring against ISO/IEC 42001 and DAMA-DMBOK2.

## Articles
- [Responsible AI Principles in Practice](https://yukon.az/blog/responsible-ai-principles): Turning responsible AI principles into controls: what each principle means operationally, the control that implements it, and the evidence that proves it was applied.
- [Enterprise AI Security: Threat Model and Controls](https://yukon.az/blog/enterprise-ai-security): A security programme for the AI estate: system inventory, risk classification, seven control families, third-party model risk, AI incident response and framework mapping.
- [AI Adoption Roadmap: From Pilot to Enterprise Scale](https://yukon.az/blog/ai-adoption-roadmap): A phased AI adoption roadmap with explicit gates between phases, the staffing each phase needs, and what has to be true before the first deployment starts.
- [Common AI Implementation Mistakes Enterprises Make](https://yukon.az/blog/ai-implementation-mistakes): Ten mistakes that keep enterprise AI projects stuck in pilot — from starting with the model to leaving identity until late — and the sequence that avoids them.
- [How HAVAA Enables Enterprise AI Transformation](https://yukon.az/blog/how-havaa-enables-ai-transformation): What HAVAA provides that a framework does not: four safety modes, human approvals, a full audit trail, database-level tenant isolation and sovereign deployment.
- [AI Automation vs Traditional Automation (RPA)](https://yukon.az/blog/ai-automation-vs-rpa): Where RPA still wins, where AI agents win, and why most enterprises need both — with a decision rule you can apply per process step and advice on an existing bot estate.
- [Choosing an Enterprise AI Platform: An Evaluation Framework](https://yukon.az/blog/choosing-an-enterprise-ai-platform): Cloud platform, open-source build or deployable product: the ten criteria that actually separate enterprise AI platforms, and how to run an evaluation that predicts production.
- [Why Context Matters in Enterprise AI: RAG, Grounding and Retrieval](https://yukon.az/blog/context-in-enterprise-ai-rag): Why retrieval quality determines enterprise AI results more than model choice, the difference between training and grounding, and how to evaluate retrieval on its own.
- [How AI Agents Automate Business Processes](https://yukon.az/blog/ai-agents-business-process-automation): How to decompose a business process for agent automation: the deterministic and probabilistic split, where the human stays, exception handling and how to measure the result.
- [AI Agent Use Cases Across Industries](https://yukon.az/blog/ai-agent-use-cases): Where enterprise AI agents actually deliver — banking, government, insurance, telecom, industry and internal functions — plus the use cases that consistently fail.
- [Building Secure AI Workflows: Auth, Secrets, Sandboxing and Audit](https://yukon.az/blog/secure-ai-workflows): The security architecture for enterprise AI workflows: identity propagation, why prompt injection cannot be fixed by prompting, secrets handling, sandboxing and audit.
- [AI Orchestration Architecture: Control Plane, Tools, Memory and State](https://yukon.az/blog/ai-orchestration-architecture): The components of an enterprise AI orchestration layer — control plane, tool boundary, memory and state, routing, policy gates and audit — and how they fail.
- [Data Lakehouse Best Practices](https://yukon.az/blog/data-lakehouse-best-practices): Practices that keep a lakehouse healthy past year one: table layout, ingestion discipline, governed layers, maintenance jobs, consumption rules and an annual review.
- [How Starburst Accelerates Analytics: Pushdown, Caching and Indexing](https://yukon.az/blog/how-starburst-accelerates-analytics): The mechanisms behind federated query performance — predicate and aggregate pushdown, parallel JDBC extraction, caching, materialised views and data skipping — and how to tune them.
- [Real-Time Analytics: Streaming, CDC and the Lakehouse](https://yukon.az/blog/real-time-analytics-streaming-cdc): How change data capture and streaming feed a lakehouse without overwhelming source systems — latency tiers, the small files trade-off, and when real-time is not worth it.
- [Lakehouse Architecture for Enterprise AI and RAG](https://yukon.az/blog/lakehouse-for-enterprise-ai-rag): Why retrieval-augmented generation depends on the data platform underneath it: where embeddings live, how access control reaches retrieval, and keeping the index current.
- [Data Lakehouse Implementation: The Challenges Nobody Warns You About](https://yukon.az/blog/data-lakehouse-implementation-challenges): Small files, wrong partitions, two kinds of catalog, schema drift, access control that stops at the bucket, and the migration that never ends — with the fix for each.
- [Building a Modern Data Platform: A Reference Architecture](https://yukon.az/blog/modern-data-platform-reference-architecture): An eight-layer reference architecture for an enterprise data platform — ingestion, storage, transformation, query, semantics, governance, orchestration and AI — with build order.
- [Why Enterprises Are Moving to Lakehouse Architecture](https://yukon.az/blog/why-enterprises-move-to-lakehouse): The five forces behind lakehouse adoption — duplicate storage cost, open table formats, AI workloads, lock-in and residency — and the cases where it is still the wrong answer.
- [GDPR vs Azerbaijani Data Protection: What Multinationals Must Reconcile](https://yukon.az/blog/gdpr-vs-azerbaijani-data-protection): Where GDPR and Azerbaijan's Law on Personal Data align, where they diverge — registration, lawful bases, transfers, rights — and how to run one control set across both.
- [Data Governance in a Trilingual Organization: AZ / EN / RU Metadata](https://yukon.az/blog/trilingual-data-governance): How to run one data catalog across Azerbaijani, English and Russian: authoritative language per term, synonym rings, transliteration, and what breaks in search and AI retrieval.
- [Central Bank of Azerbaijan: IT and Data Requirements for Financial Institutions](https://yukon.az/blog/cbar-it-data-requirements): What CBAR expects from supervised institutions on information security, data classification, outsourcing and reporting — and the data architecture that satisfies it.
- [How OvalEdge Works: Architecture, Connectors and Deployment Models](https://yukon.az/blog/how-ovaledge-works): What OvalEdge actually is, how its crawlers and lineage engine work, which deployment models exist including air-gapped, and what an implementation involves week by week.
- [Data Governance Best Practices for Regulated Industries](https://yukon.az/blog/data-governance-best-practices-regulated): Seven governance practices that hold up under supervision: evidence-first design, classification-driven enforcement, column-level lineage, access reviews and one mapped control set.
- [Common Data Governance Challenges and How to Get Past Them](https://yukon.az/blog/data-governance-challenges): The seven failure patterns that stall data governance programmes — sponsorship, scope, stewardship, adoption, definitions, tooling and outcomes — and the fix for each.
- [Where Data Quality Fits in a Data Governance Program](https://yukon.az/blog/data-quality-in-data-governance): Data quality is a discipline inside data governance, not an alternative to it. Which dimensions to measure, where rules come from, and who is accountable when one fails.
- [Data Stewardship: Roles, Responsibilities and a Working RACI](https://yukon.az/blog/data-stewardship-roles-raci): What a data steward actually does week to week, how to split the role between business and IT, and a RACI that survives contact with a real organisation.
- [Data Governance Maturity Model: Where Is Your Organization?](https://yukon.az/blog/data-governance-maturity-model): A five-level data governance maturity model, an honest self-assessment you can run in an afternoon, and what actually moves an organisation from level 2 to level 3.
- [Yukon Labs joins Microsoft for Startups](https://yukon.az/blog/microsoft-for-startups): Yukon Labs has been accepted into Microsoft for Startups, unlocking $100,000 in Azure credits, Azure AI infrastructure and enterprise tooling to accelerate HAVAA, our sovereign AI platform.
- [Deploying LLMs On-Premise: Architecture, GPU Sizing and Real Costs](https://yukon.az/blog/deploying-llms-on-premise): A practical guide to running large language models in your own data centre: the hardware you need, how to size it, what it costs against API pricing, and where deployments go wrong.
- [ISO/IEC 42001: What the AI Management System Audit Actually Asks For](https://yukon.az/blog/iso-42001-audit): A practical account of ISO/IEC 42001: what the standard requires, what auditors ask to see, how it differs from ISO 27001, and what certification really costs.
- [Starburst vs Dremio vs Databricks SQL: Choosing a Query Engine](https://yukon.az/blog/starburst-vs-dremio-vs-databricks): A practical comparison of three lakehouse query engines: federation breadth, on-premise viability, performance, catalog lock-in, cost model and governance.
- [Starburst and Trino in Azerbaijan: Federated Analytics Without Pipelines](https://yukon.az/blog/starburst-trino-azerbaijan): How Starburst and Trino let Azerbaijani banks query across core banking, warehouse and object storage in one SQL statement — on-premise, without moving data.
- [Data Fabric vs Data Mesh: Two Answers to the Same Problem](https://yukon.az/blog/data-fabric-vs-data-mesh): Data fabric is a technical answer, data mesh an organisational one, to the same problem: central data teams cannot keep up. When each works, and why most institutions need both.
- [Apache Iceberg vs Delta Lake vs Hudi: Choosing a Table Format](https://yukon.az/blog/iceberg-vs-delta-lake-vs-hudi): Three open table formats, one decision that shapes your engine choices for years. Where they differ on writes, catalogs and lock-in — and why the gap narrowed in 2026.
- [The EU AI Act: Obligations for Azerbaijani Companies Serving EU Clients](https://yukon.az/blog/eu-ai-act-azerbaijan): The AI Act reaches beyond the EU. Which Azerbaijani companies it captures, what high-risk classification actually requires, how the deadlines moved, and what to do this year.
- [Data Residency and Personal Data Law in Azerbaijan: An AI Compliance Guide](https://yukon.az/blog/data-residency-azerbaijan): What Azerbaijan's personal data regime means for AI and analytics: where data may be processed, what cross-border transfer requires, and how to design systems for it.
- [Data Catalog: Build vs Buy](https://yukon.az/blog/data-catalog-build-vs-buy): Building a catalog looks cheap because the first version is. The four capabilities that decide it, the honest three-year cost comparison, and the two cases where building is right.
- [OvalEdge vs Atlan: Catalog Comparison for Regulated Industries](https://yukon.az/blog/ovaledge-vs-atlan): Atlan leads on user experience and active metadata; OvalEdge on on-premise viability, connector depth and cost. An implementer's comparison for supervised institutions.
- [Building a Data Governance Program: A 12-Month Roadmap](https://yukon.az/blog/data-governance-roadmap): A quarter-by-quarter plan for a governance programme that survives year one: what to fund, who to staff, what to demonstrate at each gate, and what to do when it stalls.
- [What Is AI Governance? Policy, Risk and Control for Enterprise AI](https://yukon.az/blog/what-is-ai-governance): AI governance is the operating model for deciding which AI systems may exist, under what controls, with what evidence. The inventory, the risk tiers, the controls and what auditors ask.
- [Trino vs Presto: What Actually Changed and Which One to Use](https://yukon.az/blog/trino-vs-presto): Trino and Presto share an origin and diverged in 2018. What the fork actually changed, how the two projects differ now, and which one a new deployment should choose.
- [What Is Data Virtualization? Query Federation Without Copying Data](https://yukon.az/blog/what-is-data-virtualization): Data virtualization queries data where it lives instead of copying it. How pushdown works, where federation beats a pipeline, where it does not, and how to size a first deployment.
- [Air-Gapped AI: Running Enterprise AI With No Internet Egress](https://yukon.az/blog/air-gapped-ai): Air-gapped AI means every component runs with no outbound internet. What that actually requires, where sovereignty claims quietly break, and how to operate models with no egress.
- [What Is MCP (Model Context Protocol) and Why It Matters for Enterprise AI](https://yukon.az/blog/what-is-mcp): MCP is the open standard for connecting models to tools and data. How it works, what changed when it moved to the Linux Foundation, and the security questions to ask before deploying.
- [Multi-Agent Systems in the Enterprise: Architecture and Failure Modes](https://yukon.az/blog/multi-agent-systems-enterprise): When multiple AI agents are worth the complexity, the four architectures that actually work, and the failure modes — compounding error, deadlock, cost blowup — that cancel projects.
- [Data Catalog vs Data Dictionary vs Business Glossary](https://yukon.az/blog/data-catalog-vs-data-dictionary-vs-business-glossary): Three artefacts, three audiences, three failure modes. What each one is for, why organisations build the wrong one first, and how they fit together in a governance programme.
- [Business Glossary: Turning Definitions into Governed Assets](https://yukon.az/blog/business-glossary): A business glossary is where governance stops being technical. How to define terms that survive an argument, bind them to physical columns, and keep them alive after launch.
- [Metadata Management: Technical, Business and Operational Metadata](https://yukon.az/blog/metadata-management): The three kinds of metadata, why conflating them breaks governance programmes, how active metadata differs from a static inventory, and what to harvest rather than type.
- [Understanding Data Lineage: Column-Level Tracing in Practice](https://yukon.az/blog/data-lineage-explained): Data lineage traces a number in a report back to every source column that produced it. What column-level lineage really requires, where automated harvesting breaks, and how to test it.
- [What Is Data Governance? Framework, Roles and Operating Model](https://yukon.az/blog/what-is-data-governance): Data governance is an operating model, not a policy binder. The framework, the roles that make it work, the decision rights that give it teeth, and how to tell whether it exists.
- [What Is an AI Agent? Enterprise Agents vs Chatbots vs Assistants](https://yukon.az/blog/what-is-an-ai-agent): An AI agent pursues a goal across multiple steps using tools, rather than answering one question. What separates agents from chatbots and assistants, and what enterprises get wrong.
- [Data Governance in Azerbaijan: A Practical Guide for Banks and Government](https://yukon.az/blog/data-governance-azerbaijan): How data governance works in Azerbaijani banks and state institutions: the regulatory drivers, trilingual metadata, on-premise constraints and a 12-month programme that delivers.
- [Data Lakehouse Architecture: Warehouse vs Lake vs Lakehouse](https://yukon.az/blog/data-lakehouse-architecture): What a data lakehouse actually is, how open table formats made it possible, and an honest comparison against the warehouse and the lake it replaces.
- [What Is AI Orchestration? Agents, Workflows and the Enterprise Control Plane](https://yukon.az/blog/what-is-ai-orchestration): AI orchestration is the layer between your models and your business systems: routing, tool access, state, permissions and audit. Why an LLM API alone does not scale.
- [Enterprise AI in Azerbaijan: Adoption, Barriers and What Actually Works](https://yukon.az/blog/enterprise-ai-azerbaijan): What enterprise AI adoption looks like in Azerbaijani banks and government: the real barriers, the use cases that reach production, and the ones that stall.
- [Sovereign AI: Why Governments and Banks Require On-Premise Deployment](https://yukon.az/blog/sovereign-ai): Sovereign AI means running AI entirely under your own legal and physical control. What that requires technically, what it costs, and why regulated institutions have no alternative.
- [OvalEdge vs Collibra vs Alation: An Implementer's Comparison](https://yukon.az/blog/ovaledge-vs-collibra-vs-alation): An honest comparison of three data catalogs from the team that deploys them: cost, on-premise viability, lineage quality, analyst experience and time to value.
- [What Is a Data Catalog? Metadata, Lineage and Business Glossary Explained](https://yukon.az/blog/what-is-a-data-catalog): A data catalog indexes what data exists, what it means, where it came from and who owns it. How metadata, column-level lineage and the business glossary work.

## Notes

- Canonical language is English at the root; other locales are a trailing path segment.
- The company name is "Yukon Labs" — never the singular form, and never "Yukon" on its own.
- The AI platform is named "HAVAA". "Havaa AI" is an alternate spelling of the same product, not a second one; the URL segment is havaa-ai.
- The third solution is "AI orchestration". "AI governance" is the name of an assessment track, not a Yukon Labs offering.
- Solution pages describe the discipline; the platform page under each describes the product that delivers it.