Glossary

→ Data Warehouse
→ Data Lakehouse
→ Data Mesh

Business Intelligence, or BI, is a way for companies to understand their data and make better decisions. It involves collecting information from different parts of the business—like sales, finance, or customer service—and turning it into easy-to-understand reports, charts, and dashboards. BI helps people see what’s happening in the business, spot trends, and make informed choices. Instead of guessing, BI gives you a clear picture based on real numbers.

→ Generative AI
→ Natural Language Processing (NLP)
→ Large Language Model (LLM)

Artificial Intelligence, or AI, refers to the ability of machines and computer programs to perform tasks that normally require human intelligence. This includes things like learning from experience, understanding language, recognizing patterns, solving problems, and making decisions. AI is behind technologies like voice assistants (e.g., Siri or Alexa), chatbots (ChatGPT, Claude, Gemini etc.), recommendation systems (like seen in Netflix or Spotify), self-driving cars. It’s about making computers smart enough to help us in ways that feel natural and intuitive.

→ Data Governance
→ Data Literacy
→ Data Quality

Data culture is a shared organizational mindset and set of practices in which decisions, processes, and strategies are consistently informed by data. It emphasizes collaboration, trust, and transparency around data usage and insights, encouraging employees at all levels to value and leverage data for continuous improvement and innovation.

→ Data Governance
→ Data Culture
→ Data Quality

Data literacy is the ability to read, interpret, analyze, and communicate insights from data in a way that informs effective decision-making. It involves understanding how data is gathered, what it represents, and how to question or validate its findings, ultimately fostering a more informed and data-driven culture within an organization or society.

→ Data Governance
→ Data Literacy
→ Data Quality

A data catalog is a centralized, searchable repository that provides metadata and contextual information about an organization’s data assets, enabling users to discover, understand, and manage available data. It typically includes details like data ownership, classification, lineage, and usage guidelines, streamlining data discovery and fostering a more data-driven culture through better transparency and governance.

Some of the well-known data catalogue tools are: Alation, Collibra, AWS Glue Data Catalog, Informatica Data Governance and Catalog

→ Data Strategy
→ Data Catalog
→ Data Quality

Data Governance is a framework of policies, procedures, and standards that ensures data is consistent, accurate, secure, and used responsibly throughout an organization. It defines who can take what action upon what data, in what situations, and using which methods, helping to maintain data quality, compliance, and alignment with overall business objectives.

ome key aspects typically included in a data governance framework:

1. Roles & Responsibilities
   Clear ownership and accountability (e.g., data owners, stewards, custodians) is absolutely crucial for a successful Data Governance. It ensures people know who is responsible for specific data assets and governance activities.

2. Policies & Standards
   Establishes formal rules, procedures, and quality benchmarks for data creation, storage, and usage, ensuring consistency and compliance across the organization.

3. Data Quality Management
   Implements processes and criteria (like accuracy, completeness, and timeliness) to maintain high data quality and trustworthiness.

4. Security & Privacy
   Protects sensitive data through access controls, encryption, and adherence to relevant regulations (e.g., GDPR, HIPAA).

5. Data Lifecycle Management
   Governs data from creation to archival, including retention and disposal practices, ensuring optimal storage and regulatory compliance.

6. Metadata Management
   Captures information (metadata) about data (e.g., definitions, relationships, lineage), enabling efficient data discovery and understanding. Often handled via a Data Catalogue Tool.

7. Monitoring & Auditing
   Continuously tracks compliance, identifies anomalies, and evaluates the effectiveness of governance practices, enabling adjustments as needed.

→ Data Governance
→ Data Literacy
→ Data Quality

Data Strategy is a comprehensive plan that outlines how an organization will leverage data to achieve its business objectives. It aligns with the overall corporate vision, ensuring that data is collected, managed, analyzed, and used effectively. Below are seven key aspects of a solid data strategy:

• Organization Strategy: Align data initiatives with overall business goals and resource allocation.

• Use Cases & Fields of Action: Define focused data projects that address critical business challenges or opportunities.

• Data Governance: Establish rules and processes to ensure secure, consistent, and compliant data management.

• Data & Analytics Capabilities: Equip teams with the methods and technology to derive actionable insights from data.

• Talent & Culture: Foster a workforce that values, understands, and actively leverages data.

• Data Quality: Ensure accuracy, consistency, and completeness for reliable insights and decisions.

• Platform & Tools: Adopt scalable, efficient infrastructure to store, process, and analyze data effectively.

→ Generative AI
→ Natural Language Processing (NLP)
→Transformer Architecture
→ Large Language Model (LLM)

A Generative Adversarial Network (GAN) is a machine learning framework involving two neural networks—often referred to as the Generator and the Discriminator—that are trained simultaneously in a minimax game:

1. Generator (G)

   → Creates (or “generates”) new data that mimics the characteristics of real data.
   → Learns from feedback provided by the Discriminator, continually refining its ability to produce more convincing outputs.

2. Discriminator (D)

   → Evaluates whether the inputs it receives are real (from the training dataset) or generated (from the Generator).
   → Provides feedback to the Generator, indicating how closely its outputs resemble true data examples.

By competing against each other, both networks improve over time. The Generator becomes better at creating realistic data, and the Discriminator becomes more adept at detecting generated vs. real data. Although best known for image generation – such as producing photorealistic images – GANs have also been adapted for text, audio, and various other data types.

→ Generative AI
→ Natural Language Processing (NLP)
→ Generative Adversarial Network (GAN)
→ Large Language Model (LLM)

A Transformer is a neural network design that revolutionized natural language processing (NLP) by introducing a self-attention mechanism. Unlike older architectures (e.g., RNNs or LSTMs), Transformers process entire sequences in parallel, making them highly efficient and more capable of understanding long-range dependencies. Key components include:

• Self-Attention: Allows the model to weigh the relevance of different words (tokens) when generating or interpreting text.

• Multi-Head Attention: Splits the self-attention process into multiple “heads,” enabling the model to focus on different parts of the sequence simultaneously.

• Feed-Forward Layers: Fully connected layers that refine the attention outputs for better feature representation.

• Positional Encoding: Injects positional information about tokens (like word order) since the architecture does not inherently track sequence order.

Because of its parallelizable design and ability to handle context across large spans of text, the Transformer architecture powers many modern Large Language Models (LLMs), achieving state-of-the-art performance in tasks like machine translation, text summarization, and content generation.

→ Generative AI
→ Natural Language Processing (NLP)
→ Generative Adversarial Network (GAN)
→ Transformer Architecture

A Large Language Model (LLM) is a powerful Artificial Intelligence System that is able to understand and generate human-like text. They are trained on massive amounts of text data and learn to recognize patterns and relationships between words, enabling them to produce coherent text, answer questions, and even create new content. Thanks to breakthroughs in Natural Language Processing (NLP) and machine learning (ML), LLMs have significantly improved in recent years, becoming powerful tools for chatbots, content generation, automated translations, and more.

The earliest concept involves a Generator and a Discriminator, often referenced in a Generative Adversarial Network (GAN). While the concepts have evolved a lot over time, it still makes the basic principle of how an LLM works quite understandable:

1. The Generator produces text (or tokens), based on the patterns it has learned from the material it was trained on. It tries to mimic text and make it look as authentic as possible, but it cannot judge if it’s real or not. The Generator can generate huge amounts of text within seconds.
2. The Discriminator then acts as an evaluator, distinguishing between real and generated text. It provides feedback or even corrections to refine the Generator. If the discriminator detects the generators output as fake, the generator is penalized and learns to improve.

Modern Large Language Models (LLMs) typically do not use a separate adversarial Discriminator. Instead, they rely on a single model refined through self-attention mechanisms, allowing them to handle long-range dependencies in text. This approach enables highly efficient and powerful performance in natural language processing tasks. Though other architectures exist (e.g., RNNs, LSTMs), the Transformer remains the dominant choice for state-of-the-art LLMs due to its scalability and overall effectiveness.