IA EMPRESARIAL: SLMS, RAG Y USO SEGURO DE LA INTELIGENCIA ARTIFICIAL

Curso de IA empresarial enfocado en SLMs, RAG y uso seguro de la inteligencia artificial. Aprende a diseñar soluciones eficientes, privadas y aplicadas a negocio. ?? Importante: el curso se imparte íntegramente en inglés

I.
Small Language Models (SLMs): Private AI, Edge & Strategy
1.
Introduction to SLMs: Why Now?
1.1
AI Evolution: From Giant LLMs to Smart SLMs (Phi-3, Gemma 2, Llama 3.2)
1.2
Problems with Large LLMs: Cost, Latency, Privacy
1.3
Competitive Advantage: SLMs on Edge Devices, Mobile Apps, Local Intranets
2.
What Are SLMs? Key Concepts Without Math
2.1
Simple Technical Difference: Fewer Parameters = More Efficiency (e.g., 3B vs 70B)
2.2
How They Work: Distillation, Pruning, Quantization Explained Visually
2.3
Visual Comparison: SLM vs LLM in Speed, Memory, Accuracy
3.
Business Benefits: Efficiency and Scalability
3.1
Cost: Dramatically Cheaper in Inference and Deployment
3.2
Privacy: Local Models Without Sending Data to the Cloud
3.3
Speed: Millisecond Responses for Real-Time Apps
3.4
Initial Cases: Internal Chatbots, Offline Analysis, Edge Personalization
4.
How Non-Technical People CAN Use SLMs TODAY
4.1
Everyday Apps with Embedded SLMs: Mobile (Local Assistants), Office (Copilot Edge), Browsers
4.2
Accessible Platforms: Ollama Desktop, LM Studio, Mobile Apps with Embedded SLMs
4.3
Practical Examples in Slides: "Chat with Your Documents Offline", "Local Summaries", "Translator Without Internet"
4.4
Checklist: 3 Ways to Try SLMs Today Without IT or Programming
5.
Business Use Cases for SLMs
5.1
Customer Support: SLMs in Mobile Apps Without Connection
5.2
Field Service: AI on Tablets for Technicians (Offline Diagnostics)
5.3
Knowledge Workers: Local Assistants for Internal Documents/Processes
5.4
IoT/Edge: Prediction in Factories, Smart Sensors
5.5
Mini Exercise: Map 3 Processes in Your Company to Possible SLMs
6.
Limitations and When NOT to Use SLMs
6.1
Precision Gap: Complex Tasks Where LLMs Win
6.2
Fine-tuning Challenges vs Mature LLMs
6.3
Decision Framework: SLM vs LLM vs Simple Rules
7.
Non-Technical Implementation: Working with Teams
7.1
How to Present SLM Use Cases to IT/Data Teams (Expected Inputs/Outputs)
7.2
Common Platforms (Hugging Face, ONNX, Ollama) - High Level
7.3
Real Costs: Minimum Hardware, Cloud vs On-Premise
7.4
Business Case Template for SLMs
8.
Future of SLMs and Next Steps
8.1
2026 Roadmap: Multimodal SLMs, Local Agents
8.2
Pioneer Companies: Microsoft Phi, Google Gemma Cases in Production
8.3
Checklist: Evaluate if Your Next AI Project Needs an SLM
8.4
Final Workshop: From 3 Generic Ideas, Select the Best One for SLM
II.
RAG for Enterprise Knowledge: Making AI work with your company's data
1.
Introduction to RAG: The Missing Piece in Enterprise AI
1.1
The Problem: Why ChatGPT Doesn't Know Your Company
1.2
What Goes Wrong: Hallucinations, Outdated Info, Generic Answers
1.3
The Solution: RAG Brings Your Documents to AI
2.
What is RAG? Core Concepts Without the Math
2.1
The Two-Step Process: Retrieve, Then Generate
2.2
How RAG Finds Relevant Information (Embeddings Explained Simply)
2.3
Why RAG Reduces Hallucinations and Improves Accuracy
3.
Business Benefits: Accuracy, Control, and Up-to-Date Knowledge
3.1
Accuracy: Answers Grounded in Your Documents
3.2
Control: You Decide What Data AI Can Access
3.3
Always Current: Update Documents, Not Models
3.4
Traceability: Citations and Source References
4.
How Non-Technical People Can Use RAG TODAY
4.1
Category 1: "Personal RAG" (Ad-hoc Tools)
4.2
Category 2: "Corporate Search Engines" (Enterprise Search)
4.3
Category 3: "RAG Builders" (Low-Code / Custom)
4.4
Decision Framework: Which Category Do You Need?
5.
Enterprise Use Cases for RAG
5.1
Internal Knowledge Bases: HR, IT, Operations Manuals
5.2
Customer Support: Technical Documentation and Product Guides
5.3
Research and Analysis: Legal, Compliance, Market Research
5.4
Sales Enablement: RFP Responses, Competitive Intelligence
5.5
Mini Exercise: Identify 3 Document-Heavy Processes in Your Company
6.
Limitations and When RAG Isn't Enough
6.1
Retrieval Quality: Garbage In, Garbage Out
6.2
Context Window Limits: Too Much or Too Little Information
6.3
When to Use RAG vs Fine-Tuning vs Hybrid Approaches
7.
Working with Teams: RAG Implementation Without Coding
7.1
How to Specify RAG Requirements to IT Teams
7.2
Document Preparation: What Makes Good RAG Data
7.3
Common Platforms: Pinecone, Weaviate, LangChain (High Level)
7.4
Rag Closing
III.
AI Accuracy & Guardrails: Stop Hallucinations & Spot Errors
1.
Why AI Gets Things Wrong
1.1
How generative models work at a high level (statistical patterns, not logic)
1.2
Hallucinations: what they are and why they occur (invented data, references, facts)
1.3
Other common errors: biases, vague answers, misunderstanding context
2.
Spotting AI Errors in Real Time
2.1
Warning signs in text: overly confident tone, lack of sources, contradictions
2.2
Warning signs in data/numbers: impossible calculations, strange dates, dubious references
2.3
Practical examples: “good response” vs “hallucination” side by side
3.
Questioning & Probing Techniques
3.1
Verification questions: “How do you know that?”, “Show your steps”, “What alternatives exist?”
3.2
Asking for self-review: “Check your response for errors”, “Critique your own argument”
3.3
Effective patterns: Devil’s Advocate, Multi-View, Step-By-Step
4.
Simple Guardrails You Can Ask the AI to Apply
4.1
Structural guardrails
4.2
Verification guardrails
4.3
Consistency guardrails
4.4
Guardrails walkthrough
4.5
Guardrail prompt templates
5.
Understanding AI Limits
5.1
Why AI sounds confident
5.2
Structural limitations
5.3
When not to trust AI
6.
Safe AI Use at Work
6.1
Golden rules for safe use
6.2
AI + humans + sources
6.3
Final error-fixing workshop
7.
Bonus: Future AI Risks
7.1
Goal-driven AI risks
7.2
More autonomous AI
7.3
Why humans stay in the loop

-Comprender el uso de la IA empresarial y su aplicación en entornos de negocio.
-Diferenciar entre modelos de lenguaje grandes (LLMs) y modelos eficientes (SLMs).
-Identificar casos de uso de SLMs en entornos empresariales con foco en eficiencia y privacidad.
-Entender el funcionamiento de los sistemas RAG y su aplicación con datos internos de la empresa.
-Conocer los elementos clave de un sistema RAG, incluyendo preparación de datos y recuperación de información.
-Aplicar técnicas para reducir errores y alucinaciones en sistemas de inteligencia artificial.
-Utilizar prompting estructurado y estrategias de verificación para mejorar la fiabilidad de las respuestas.
-Definir un uso seguro de la inteligencia artificial en procesos y flujos de trabajo empresariales.
Importante: el curso se imparte íntegramente en inglés, aunque el soporte y la información del curso se presentan en español.