1. Beyond Blueprints: Using AI to Automate Architecture Definitions and Documentation 🎯

Speaker: Mario Lopez

The magic formula for AI-powered architecture documentation:

• Gather context → Define goals → Craft precise prompts → Review & iterate
• "A good architecture definition evolves with the applications"

Key frameworks:

• C4 Model
• 4+1 View Model
• Viewpoint & Perspectives
• Just Enough Documentation

---

2. From User Stories to Production Using AI ⚠️

Speaker: Diego Barahona

Main insight: "AI writes code in seconds but it rarely works in production."

• The issue is not capability—it's direction
• The missing link: workflow and alignment (input → build → test → review → deploy)
• AI needs structure like CI/CD enforces discipline
• Predictable results come from alignment
• Solution: Agent-OS - standards, product, specs

---

3. MongoDB - Demystify NoSQL 🗄️

Speaker: Luis Cordoba

Quote: "Data used to live in tables. Now it lives everywhere - in tweets, NFTs, LLM Embeddings, and sensor streams. And yet... we still try to fit it into rows and columns."

Key Insights:

• Without Data, there is no AI!

MongoDB Pros:

• Flexible Schema (with Schema Validation when needed)
• Horizontal Scalability (sharding)
• Excellent Developer Experience (JSON)

MongoDB Cons:

• High Memory Usage (Higher RAM)
• Not Ideal for Complex Transactions
• Consistency Trade-offs

Takeaways:

• Learn what NoSQL IS capable of doing
• Learn what NoSQL is NOT capable
• Don't force patterns on technologies that were not made for them
• Stop the MongoDB Hate! 😄

---

4. From Regulation to Innovation: Implementing AI in Finance and Healthcare ✋

Speaker: Rodrigo Garcia

Main principle: "Every decision needs a human who is accountable"

Key Insights:

The gap between success and failure is bridgeable with the right approach:

• AI is transformative but requires strategic, measured deployment
• Assess data readiness honestly (77% rate theirs as poor)
• Start with back-office, measurable processes
• Build compliance frameworks NOW, not later
• Focus on business outcomes, not AI metrics

Human-AI Boundary by Risk Level:

Risk Level	Approach	Examples
Low Risk	Full automation	Routine transactions, data classification
High Risk	Human-in-loop + AI support	Major credit decisions, diagnoses, fraud detection
Critical	Human-led + AI assistance	Life-critical decisions

---

5. Current State of AI 💡

Panel: Freddy Montes, Alfredo Bonilla, Mariano Alvarez

The Big Question:

Will AI replace your job?

• YES, if you don't adapt
• NO, if you start NOW

Key Insights:

• Best use case mentioned: More time to play guitar 🎸 (half-joking, but actually true—AI handles the grunt work)
• Not using AI today = still programming with punched cards in 2025
• Tools worth exploring: Lovable for rapid prototyping

---

Resources by Talk

Talk 1 - Architecture with AI:

🔗 C4 Model

Talk 2 - User Stories to Production:

🔗 Agent-OS

Talk 3 - MongoDB/NoSQL:

🔗 MongoDB University

Talk 5 - Current State of AI:

🔗 Lovable (AI Prototyping)

---

Acknowledgments

Thanks to Telescoped community, Blockchain Jungle, and the amazing speakers:

• Mario Lopez
• Diego Barahona
• Luis Cordoba
• Rodrigo Garcia
• Freddy Montes
• Alfredo Bonilla
• Mariano Alvarez

---

Tags

#AI #SoftwareArchitecture #MachineLearning #MongoDB #NoSQL #BlockchainJungle #CostaRica

---

Author: Nayib Sarmiento
Event: Blockchain Jungle + Telescoped AI Workshops
Date: November 2025
Location: Costa Rica