9:35 AM — 10:20 AMPlenary
EVENT SCHEDULE // OCT 9-10, 2026
Two days of talks, workshops & community
Explore the full schedule. Filter by day, track and topic, and plan your two days at DevFest Santo Domingo.
The two talk tracks run in parallel; workshops follow their own schedule.
Day 1 — Fri, Oct 9
Track 1Salón Principal
Track 2Salón B
WorkshopsLaboratorio
9:15 AM
⏸
OPENING
9:35 AM
10:20 AM
☕
BREAK
10:40 AM
10:40 AM — 11:20 AMSalón Principal
Track 1
Learn how we built an infrastructure that handles 300 million daily requests across 45 different blockchains in US/EU/Japan on an efficient budget using k8s, haproxy, docker and cloudflare to set up a global DNS where every change goes through a memory brain without breaking anything in the process Key points: Infrastructure; AI; Blockchain; Web3; DNS
CloudBlockchainDevOps
Lowell Abbott
Co-Founder nodefleet.net
10:40 AM — 11:20 AMSalón B
Track 2
This session presents, in practical terms, how to design enterprise automations capable of executing complete processes and reserving human intervention for exceptional situations. Using the operation of a Dominican courier company as the common thread, we will learn to break a process down into events, states, rules, data sources, tools and decisions. This separation will make it possible to determine which tasks should be solved through traditional automation and which can benefit from artificial intelligence's capacity for interpretation and reasoning. We will see how to turn emails, documents and unstructured messages into reliable data; how to use function calling to connect AI models with enterprise tools; and how to coordinate modular workflows through n8n, PostgreSQL, APIs and webhooks. We will also address state handling, transition validation, idempotency and the separation between probabilistic decisions and deterministic operations. The session will explain how to design error-resistant flows through retries, traceability, observability, least privilege, validations and human intervention mechanisms. It will show how to prevent the model from directly performing financial calculations or critical modifications without verifiable controls. During the demo we will build and run a representative sequence that starts with an external event and ends with recording a transaction, its invoicing, notification and next operational action. The goal will be to understand the architectural pattern, not just the business case used. By the end, attendees will have a methodology for identifying automatable processes, defining the tools available to an agent, setting limits of autonomy and evolving from isolated assistants toward autonomous enterprise workflows. The same approach can be applied to logistics, sales, customer service, finance and administrative operations. The central idea of the session is that, when systems can understand events, query information and execute actions safely, interfaces stop being a permanent requirement and become an optional mechanism for supervision and control.
AI/MLProduct & Startups
Wagner Luciano
CEO | Founder | Software Architect | AI & Logistics Technology Strategist
10:40 AM — 12:05 PMLaboratorio
Workshops
Explore what Gemma 4 means for the future of open-source AI. In this session, we’ll break down its capabilities, practical use cases, and the growing opportunity for developers and organizations to build, customize, and deploy powerful AI systems with greater control, transparency, and accessibility. Key points: The importance of open source, how Gemma works, the advantages of a local model
AI/ML
Carlos Alarcón
CTO at Quix | GDE | MVP | Youtuber
11:25 AM
11:25 AM — 12:05 PMSalón Principal
Track 1
It is common to see brilliant tech projects stall in their beta phase, not because of failures in the code, but because of the lack of a "human connection protocol". In the entrepreneurial ecosystem, knowing how to code, analyze data or configure servers is only half the job; the other half happens away from the keyboard, at negotiating tables and industry events. "Compiling Your StartUp: From the Terminal to Closing Deals" tackles head-on the biggest "bug" of technical founders: the fear of sales and of business protocol. During this session, we will break down the human hardware and software you need to build a company from scratch. You will learn why a firm handshake and the strategic use of business cards (physical and digital) are as crucial as clean code. Beyond attitude, the talk provides an operational roadmap. We will explore how to optimize your "bandwidth" through time management, the importance of projecting a professional image (*smart casual*) that builds trust, and why taking the step toward legal and tax formalization is your company's true deployment to production. If you are ready to step out of the IDE (*Integrated Development Environment*), network with impact and turn your ideas into formal, profitable and scalable businesses, this talk will hand you the algorithm to pull it off. Key points: Startups; Entrepreneurship; Tech; IT; Entrepreneur
Product & StartupsCareer
José Aquino Cepeda
José M. Aquino, MSc
11:25 AM — 12:05 PMSalón B
Track 2
Tools change. Technologies change. Principles remain. And at a time when AI can generate test cases, data, automation and even analyze results, the Tester's real value lies less and less in running tests and more in knowing what to test, why to test it, where to look for risk and how to interpret what they find. Key points: Apply the 7 Testing Principles to make better decisions about what, when and how much to test.; Prioritize testing by risk and context, understanding why exhaustive Testing is impossible.; Identify how automation can create a false sense of security when tests do not evolve alongside the product.; Integrate AI into Testing activities without replacing critical thinking, risk analysis and the Tester's judgment.; Assess quality beyond “the tests passed”, taking into account real user needs and business value.
Testing / QA
Victor Leonardo
Operations Director @ Smart Testing | Software Testing Specialist and Instructor
12:05 PM
🍽
LUNCH
1:05 PM
1:05 PM — 1:45 PMSalón Principal
Track 1
Who do you become when AI stops being a chatbot and turns into an extension of your brain? What happens when, all of a sudden, you have tools to export the systems that for decades existed only in your head? In this talk I am going to tell you how I ended up throwing myself —with a fork, optimism and a fair amount of recklessness— into the shark tank that is the world of public tenders. How you build a mental system to navigate thousands of pages of technical documentation, negotiate with manufacturers on the other side of the world, develop commercial intelligence, win almost 40 million dollars in contracts... and also lose 2 million dollars over a single form, because in this business mistakes are paid for dearly. But the real story is not about tenders. It is about discovering that what truly creates value is not the documents, or the emails, or the Excel files. The real asset is the decision-making system a professional builds over years of experience. When I understood that, I stopped asking myself how to use artificial intelligence. I started asking myself how to teach artificial intelligence to think like me. That question ended up turning into an ecosystem of tools that today let me research markets, build commercial intelligence, automate processes and turn professional experience into scalable products. This is not a talk about prompts. It is a talk about how a professional can transform decades of tacit knowledge into measurable, repeatable and scalable systems. Because AI does not replace experience. AI multiplies those who know exactly how they think. Key points: How to identify the value of your experience; How to turn mental processes into a system; Why AI multiplies but does not replace human judgment; Million-dollar mistakes and how to avoid multiplying the chaos
CareerAI/ML
Maria Shrayber
Founder, GTM, RevOps, Partnerships, Ecosystems
1:05 PM — 1:45 PMSalón B
Track 2
As agentic AI matures, the first hard design decision is not which model to use, but how your agents coordinate. Google's Agent Development Kit (ADK) offers two very different answers: dynamic multi-agent routing, where LLM agents decide who to transfer control to, and structured workflow loops (LoopAgent), where execution follows a fixed sequence based on shared state. Using a deliberately deterministic domain — a multi-agent math expression evaluator — this talk makes the trade-offs visible instead of leaving them up in the air. We build both patterns and then dig into three failure modes that only show up once you have it running: runaway tool-call recursion, operator precedence breaking under fixed sequencing, and loops that end silently without returning the result. Along the way we look at nested sub-agents and error boundaries (a division-by-zero guard) and a log-based explainability timeline to trace the transitions between agents. You leave with a concrete decision framework for when dynamic flexibility beats deterministic sequencing — and when it quietly costs you latency, tokens and sanity. Key points: When to choose dynamic multi-agent routing and when to choose a workflow loop (LoopAgent) in Google ADK: a decision framework based on the nature of the task, not on what is fashionable.; Why multi-agent orchestration charges a "coordination tax": every transfer of control is an LLM call that computes nothing, and that cost scales with the complexity of the problem —; How to instrument ADK agents to measure latency, tokens and real cost per run, separating useful work from coordination overhead.; Three failure modes that only show up in production: runaway tool-call recursion, operator precedence breaking under fixed sequencing, and loops that end silently without retur; Why predictability (low variance in cost and latency) is often worth more than the average when planning for production, and which pattern gives it to you.
AI/ML
Julio Quinteros P.
Google Developer Expert | AI and GCP Specialist
1:05 PM — 2:30 PMLaboratorio
Workshops
AI tools have already changed how we design. What almost nobody is discussing is how they are going to change what we design. In this follow-up to the previous talk, we go beyond the current state of the tools to ask the uncomfortable questions: which parts of the craft are about to stop existing? Which skills go from “nice to have” to being the real difference between a designer and a Figma operator? And how do you design an interface when the user no longer navigates, but delegates? We will talk about the surfaces that do not exist yet and that someone is going to have to design: interfaces for supervising agents, UIs generated at runtime, systems where design is not a screen but a behavior, and the new problems of trust, permissions and traceability that come with them. No hype and no doom. An honest read of the field from practice, for designers who want to still be relevant in five years.
AI/MLUI/UX
Claudio Henriquez
CoFounder @ thraive.do
1:50 PM
1:50 PM — 2:30 PMSalón Principal
Track 1
For the first time in history, a single person with judgment + AI + modern tools can build, distribute and operate an entire company at scale — and that changes who can start a business and how businesses get started. In this talk I am going to share my experience and the thesis I hold about how the world is changing. Key points: AI; Entrepreneurship
AI/MLProduct & Startups
Leonardo Jimenez
Founder Enrankeo/Rektora
1:50 PM — 2:30 PMSalón B
Track 2
KMP promises "write once, run everywhere", but in production the real question is not whether to adopt Kotlin Multiplatform, but what to share and what not to. In this talk we lay out a conservative, proven approach: Kotlin does the heavy lifting (domain, data, networking, state) and Swift limits itself to drawing the UI, without giving up the native experience. We will see how to structure the project with a Clean Architecture familiar to any Android dev (domain, core, data, shared), how SKIE turns Flow and coroutines into something idiomatic for Swift, when to use expect/actual without abusing it, and how to package and distribute the iOS build (XCFramework + Swift Package Manager) without losing your weekend. It closes with a checklist of 10 best practices to take straight to your next project. Key points: What to share between Android and iOS and what to keep 100% native.; How to structure a KMP project with Clean Architecture (domain, core, data, shared); Why to adopt SKIE from day one.; When to use expect/actual and when to avoid it.; How to package and distribute the iOS build without CocoaPods.
Mobile
Aneudy Vargas Almanzar
Senior Mobile Developer
2:30 PM
☕
BREAK
2:50 PM
2:50 PM — 3:30 PMPlenary
3:35 PM
3:35 PM — 4:15 PMSalón Principal
Track 1
How we used AI to run the backoffice of a distributed unconference with 70+ sessions and 2,500+ attendees in under 2 months, and what we learned for the next edition.
AI/MLCommunity
Jorge Vargas
Managing Partner Venture.do
3:35 PM — 4:15 PMSalón B
Track 2
Many SaaS companies invest time and money in campaigns that generate visits but not customers. In this talk we will go through a practical framework for building a marketing strategy focused on user activation and retention. With real tactics from content, community, SEO, partnerships, DevRel and product-led growth that can be implemented from the earliest stages through to growing companies Key points: How to identify the ICP (Ideal Customer Profile) you need; How to create messages that connect with real problems and with people; Which channels work best depending on the stage of the SaaS; How to measure the full funnel: awareness, signup, activation and retention; How to use content and community to generate sustainable demand
MarketingProduct & Startups
Yuri Santana
Developer Relations/Engineering/Product @ Clerk
3:35 PM — 5:00 PMLaboratorio
Workshops
2026 has become the era of orchestration; having one agent is no longer enough. What you need is a group of agents that can solve complex tasks. This workshop will show how to organize sub-agents as Agents as tools. To tackle complex problems. We will also look at memory handling and the different responsibilities that turn memory into a layer of cognition.
AI/ML
Carmelo Cueto
Systems Architect
4:20 PM
4:20 PM — 5:00 PMSalón Principal
Track 1
What does it take to go from an isolated prompt to a system of agents that collaborates in your development cycle? In this talk we adopt the agentic mindset —decompose, delegate, review and approve— and put it into practice by building, with Google ADK, a ReviewerAgent that analyzes code and produces structured findings with Pydantic, and then chaining it to a TestAgent through SequentialAgent to form a pipeline that generates test cases ready for CI/CD. We close with orchestration patterns, real-world challenges (predictability, stability, operations) and how to scale from two agents to N without rewriting the pipeline. Key points: adk for designing agents in the SDLC
AI/MLDevOps
Adrian Catalan
GDE and Innovation Lab Director
4:20 PM — 5:00 PMSalón B
Track 2
Frameworks change, languages evolve, and tools get replaced — but reputation stays. In a market where thousands of professionals compete for the same opportunities, personal branding has become a strategic asset. This talk explores how to build an authentic professional reputation from within the tech ecosystem: what to communicate, where to do it, how to build trust, and how to stay consistent between what you do, what you share, and how you're perceived. We'll look at common mistakes, positioning strategies, and real examples of tech professionals who turned visibility into job opportunities, partnerships, investment, and leadership within their communities. Key points: Practical channels and formats to build visibility with consistency; Practical channels and formats to build visibility with consistency; How to build trust with peers, recruiters, and communities online; Common personal branding mistakes tech professionals make
CareerCommunity
Xiara Mahovy Paulino Frías
Social Communication & Digital Media Specialist | Strategy & Reputation
Day 2 — Sat, Oct 10
Track 1Salón Principal
Track 2Salón B
WorkshopsLaboratorio
9:00 AM
9:00 AM — 9:45 AMPlenary
"Speed illusion", "You are not faster, you are more productive at things that do not matter". There is something worse than speed. The instinct to smell that something is wrong in a piece of code was built by writing code, reviewing it and getting it wrong. If you stopped doing all three, that instinct is atrophying right now and there is no metric to warn you. 44% of developers write less than 10% of their code by hand. Only 48% consistently verify what AI hands them before committing. This talk is about how to measure your real speed instead of feeling it, which skills are rusting, and what to practice deliberately so you do not wake up in two years as someone who approves code they no longer understand. Key points: What to practice deliberately so you do not lose the instinct, and how often.
CareerAI/ML
Noe Branagan
AI Babysitter
9:50 AM
9:50 AM — 10:30 AMSalón Principal
Track 1
The real challenge of Artificial Intelligence today is not writing a good prompt, but building autonomous systems that execute code, interact with databases and make decisions in a safe, predictable and scalable way. In this technical session we will take theory into real practice. We will analyze how to design the architecture of an ecosystem of autonomous agents using Go, taking advantage of its high performance and concurrency, together with the Agent Development Kit (ADK). We will not stop at conceptual slides. We will build real agents live, package them into containers and use Docker AI's capabilities for environment isolation, secure context orchestration and lifecycle governance of AI workloads. If you want to move your AI initiatives from simple local scripts toward production-ready enterprise infrastructure, this session will give you the definitive blueprint. Key points: Go; Containers; AI agents; AI governance; AI architecture
AI/MLCloudDevOps
Victor S. Recio
KubeAstronaut 🚀 | Docker Captain 🐳 | Microsoft MVP 🏆 | ☁️ Cloud & Software Architect | 🏢 ex-Microsoft | 🎓Educator & Speaker 📢
9:50 AM — 10:30 AMSalón B
Track 2
Imagine a future where cloud infrastructure governs, optimizes and repairs itself using a combination of programmable economic systems and artificial intelligence. In this vision, resources such as compute, storage and bandwidth are not simply allocated statically, but dynamically negotiated, traded and optimized by autonomous agents operating in a decentralized market. These agents, deployed with CI/CD pipelines on Google Cloud, would use advanced AI techniques such as reinforcement learning, evolutionary game theory and generative models to: Predict future demand and adjust supply in real time Discover new pricing and coalition-formation strategies Automatically detect and respond to anomalies and failures Generate new network architectures and topologies optimized for cost-performance Evolve and improve themselves through mutation and artificial selection Key points: Predict future demand and adjust supply in real time; AIDiscover new pricing and coalition-formation strategies; Generate new network architectures and topologies optimized for cost-performance; Vertex and AI model training
AI/MLDevOpsBlockchain
Ayesha Yege
Scale and Optimize Engineer - Palo Alto Networks
9:50 AM — 10:30 AMLaboratorio
Workshops
In a crisis, the core problem is not the lack of data, but its asymmetry and an excess of noise. This session breaks down the engineering behind Google Crisis Resilience, showing how code structures information to save lives through systems such as Android Earthquake Alerts, Flood Hub and SOS Alerts. We will look at the platform through three architectural pillars: Distributed Telemetry: How the signal from millions of smartphone accelerometers is ingested and processed to create the largest earthquake detection network in the world. Predictive Modeling (ML): The application of Machine Learning algorithms and satellite imagery to anticipate flood trajectories before they happen. Deployment at Scale: The integration of APIs, critical data routing and the design of highly legible interfaces (UX) to distribute mass alerts when physical infrastructure collapses. This session will show how cloud computing and large-scale data processing become the fastest and most effective humanitarian tool at our disposal. Key points: Predictive modeling; Google Crisis Resilience tools
CloudCommunityAI/ML+1
Alberto Castillo
Tech Lead in Crisis Contexts | GDG Organizer | Google Gen AI Alumni & AWS AI Early Adopter
10:30 AM
☕
BREAK
10:45 AM
10:45 AM — 11:25 AMSalón Principal
Track 1
AI is making it possible to build software in hours instead of weeks. But when that prototype reaches production, the real challenges show up: architecture, maintainability, testing, security, observability and technical debt. In this talk we will explore how to take advantage of AI tools and vibe coding without sacrificing good engineering practices. We will go through a practical workflow for moving from an idea to a system ready to grow, using architecture documents, PRDs, task-based planning and AI assistants as part of the development process. If you are already using AI to write code, or you are thinking about bringing it into your team, you will discover how to speed up development without compromising the quality or the scalability of your applications.
AI/MLDevOps
Edgar Garcia
CTO @backsource.ai & @Caribium.ai
10:45 AM — 11:25 AMSalón B
Track 2
We will explore how to design solutions with Gemma by combining local execution on the edge with orchestration in services. We will review code, implementation challenges and lessons learned building productive tools on Android and the backend.
AI/MLMobileAdrian Catalan
GDE and Innovation Lab Director
10:45 AM — 12:10 PMLaboratorio
Workshops
Is Artificial Intelligence coming to replace us, or can we turn it into an ally? In this hands-on workshop you will discover, with no technical background required, how you can use AI to boost your skills, make your day-to-day tasks easier and turn your ideas into solutions. On top of that, we will work in teams to create an idea using AI tools and present it in a 3-minute pitch. The goal: learn by doing and discover everything you can achieve when you put AI to work for you. 🤖✨
AI/MLProduct & Startups
Leomaris Reyes
Android Dev
11:30 AM
11:30 AM — 12:10 PMSalón Principal
Track 1
Artificial intelligence is changing one of the fundamental rules of software development: writing code takes less and less time. But that does not mean we are building better products. When an AI agent can generate code, tests and components in minutes, the real bottleneck moves somewhere else: our ability to define precisely what we should build, why, and how we will know it works. In this session we will explore Spec-Driven Development (SDD) as a natural evolution of agile practices in teams that are starting to work with AI agents. We will see how to move from user stories designed to coordinate people to specifications capable of conveying business intent, rules, constraints and verifiable criteria to humans and agents. And we will discuss a question that may redefine the next few years of Business Agility: If AI can write the code, where is the real work of Product Owners, developers and agile teams now? It is not about developing faster. It is about shortening the distance between a business idea and working software in the customer's hands. Key points: Code is no longer the bottleneck; Agile was designed for a different world; Prompting ≠ Spec-Driven Development; From User Stories to SpecificationsThe SPEC becomes a new unit of work
AI/MLProduct & Startups
Carlos Paulino
CCTO & founder cobalto consulting / EraNova technologies
11:30 AM — 12:10 PMSalón B
Track 2
This talk translates the world of marketing into the language developers already speak: systems, automation and data. Most developers believe three things about marketing: that it is "posting on social media", that it is "hitting the promote button", and that it is a creative talent they do not have. All three are false — and that confusion is exactly why technically superior products die in anonymity while mediocre products turn into businesses. This talk takes the mindset a developer already masters — thinking in systems, data flows, conditions and automation — and uses it as a bridge to explain what digital marketing really is today. We go through one contrast after another between "what you thought it was" and "what it actually is": from social media to the complete digital ecosystem; from the improvised post to a strategy based on buyer persona psychology and awareness levels (TOFU / MOFU / BOFU); from "promoting" to advanced segmentation; and from manual work to marketing as an automated system with funnels, AI flows in n8n and agents that nurture customers 24/7. Attendees leave with a concrete shift in mindset: they stop seeing marketing as a creative black box and start seeing it as a system they can design, measure and automate — with the same mental tools they use to code. Key points: A clear mental map of what real digital marketing is made of; The awareness-levels model and the TOFU/MOFU/BOFU funnel — and how it maps to a data pipeline they already understand.; Why segmentation and audience analysis decide whether your budget brings customers or just curious onlookers.; How modern marketing is automated with funnels, AI flows (n8n) and agents — territory where a dev has an advantage.; A first actionable step to start getting known without becoming a "marketer".
MarketingData
Daviana Nuñez
Digital Marketing Strategist / Paid Media Specialist @Couplemedia.marketing
12:10 PM
🍽
LUNCH
1:00 PM
1:00 PM — 1:40 PMSalón Principal
Track 1
AI is not only changing the way we write code. It is also turning a growing share of “knowledge work” into a new surface for engineering. Processes that used to depend on tacit knowledge, manual coordination and scattered decisions can now be made explicit, turned into systems and improved on the basis of evidence. For software engineers, this is an inflection point. The same capabilities we learned in order to build software, such as decomposing problems, modeling systems, automating, testing and learning, can now be applied far beyond code. Drawing on experiences creating products, automations and businesses with AI, this talk explores what it means to move from executing requirements to applying engineering to turn knowledge, experience and ideas into outcomes. We will also talk about the opportunities and the limits of this transformation, the role of human judgment, and how software engineers can leverage what they already know in order to create with greater autonomy. Key points: Understand why a growing share of “knowledge work” is becoming a new surface for engineering.; Recognize processes, decisions and problems that can be made explicit, turned into systems and improved from evidence.; Distinguish between writing code and designing a complete system capable of producing real outcomes.; Understand the role of human judgment and the limits we must consider when bringing AI into real processes.; Apply software engineering capabilities to create products, automations and outcomes with greater autonomy.
CareerProduct & Startups
Wilson Reyes
Co-Founder: Caribium.ai & StriveUp.ai
1:00 PM — 1:40 PMSalón B
Track 2
La audiencia entenderá qué significa realmente ejecutar inteligencia artificial de forma local y cuándo tiene sentido hacerlo en lugar de depender completamente de servicios en la nube. Al terminar la charla podrán distinguir entre SLMs, LLMs y frontier models, entender por qué no todas las tareas necesitan el modelo más grande disponible y conocer los factores principales que determinan qué modelos pueden ejecutar en su propio hardware. También aprenderán conceptos prácticos como RAM, VRAM, parámetros, cuantización e inferencia, y verán cómo una computadora, workstation o Home Lab puede convertirse en infraestructura para ejecutar modelos de IA, aplicaciones y agentes.
AI/MLDevOps
Harys Vizcaino
Ingeniero en software y creador de contenido sobre IA y tecnología @harysvizcaino
1:00 PM — 1:40 PMLaboratorio
Workshops
Modern Android Development (MAD) practices revolutionized the way we build mobile applications. Now, with the arrival of small language models (SLMs) that run directly on the phone's processor, Artificial Intelligence is integrated as one more native component of the Android ecosystem. In this 100% hands-on workshop, you will learn to apply MAD Skills patterns to integrate Gemini Nano (Google AI Edge) into a modern, private and functional Android application with no internet connection required.
MobileAI/ML
Edgardo Panchana
Mobile application development
1:45 PM
1:45 PM — 2:25 PMSalón Principal
Track 1
AI access isn't the differentiator anymore. Everyone's renting the same "horse" (the model). This talk breaks down where real advantage comes from instead: the Horse (the model itself, how it reasons, where it breaks, how to push it), the Harness (the agents, tools, workflows, and guardrails you build and own), and the Hay (the proprietary data that keeps making both better over time). Héctor will walk through this using heva, the AI native, cross-border healthcare platform he's building, where agents have to work across languages, countries, providers, patients, scheduling, and payments in the real world. That approach took heva from about $100K to over $1M in ARR in roughly five months, while raising more than $10M in venture capital. The talk ends on a question for developers and entrepreneurs in the Dominican Republic: once world class models are commoditized, where does your edge actually come from? Understand the horse. Own the harness. Grow the hay. No AI/ML background needed. This is framework first, built from real production lessons, not theory.
AI/MLProduct & Startups
Hector Terrero
CEO & Co-Founder @ heva
1:45 PM — 2:25 PMSalón B
Track 2
Interfaces generated dynamically by AI make it possible to create personalized experiences, but they can also cause unexpected layout shifts, heavy client-side load and rendering delays. In this session we will explore how to optimize a generative UI through rendering strategies, streaming, DOM reuse and efficient state handling, while keeping experiences fast, fluid and accessible. Key points: Identify the performance challenges of generative UI.; Reduce layout shifts and client-side load.; Keep dynamic interfaces fast and accessible.
AI/MLUI/UXWeb
Vanessa Marely Aristizabal Angel
Google Developer Expert in Angular and Web Technologies
1:45 PM — 2:25 PMLaboratorio
Workshops
In a world where Artificial Intelligence can generate and run more and more tests, a fundamental question arises: how do we decide what to test first? We will explore how Risk-Based Testing helps us identify what really matters, prioritize our efforts and make better decisions that deliver business value, beyond the number of tests executed.
Testing / QAAI/MLVictor Leonardo
Operations Director @ Smart Testing | Software Testing Specialist and Instructor
2:25 PM
☕
BREAK
2:40 PM
2:40 PM — 3:10 PMSalón Principal
Track 1
What if one of the most important pieces of our professional stack were not a technological one? Careers, startups and projects rarely move forward on technical knowledge alone. They also move forward because we find people to learn with, collaborate with, build with, sell to, hire, connect ideas with or open new possibilities with. Yet we frequently reduce networking to exchanging contacts, adding people on LinkedIn or attending events without a strategy for turning those encounters into lasting relationships. In this talk we will explore networking as human infrastructure for growth, not as a transactional practice. Drawing on real experiences from a career that has spanned advertising, technology, entrepreneurship, public innovation and AI governance, we will see how a conversation can become, years later, an opportunity, a project or a new community. We will present a practical framework based on five elements —Context, Curiosity, Contribution, Continuity and Collaboration— for building professional relationships based on trust and mutual benefit. Finally, we will build a small Relationship Stack with Google tools: Contacts to preserve context, Calendar and Tasks to follow up, Gmail and Drive to maintain continuity, Sheets as a lightweight tracker and Gemini to reduce the load of organizing and retrieving information. Technology can help us remember and follow up. Trust is still something we have to build ourselves.
CareerCommunity
Abel López
Founder @ Envision Innovation Labs | Strategy, innovation and ecosystem building
2:40 PM — 3:10 PMSalón B
Track 2
LLMs stopped being simple conversational interfaces. Today we connect them to APIs, databases, files, browsers, repositories and external tools to build agents capable of executing real actions. In this talk we will explore what happens when an LLM receives instructions from multiple sources, uses tools through tool calling or MCP, and on top of that has permissions to interact with external systems. We will see how attacks such as prompt injection and indirect prompt injection can manipulate an agent's behavior and trigger actions the developer never intended to authorize. We will analyze a typical AI agent architecture and see where the main trust boundaries appear between the user, the model, the context, the tools, MCP servers, APIs and external data. We will also review controls that can be applied from development onward: least privilege, parameter validation, tool allowlisting, credential separation, approval gates, human-in-the-loop, logging and execution limits. The goal is to understand that an agent's security cannot rely on a system prompt alone. If an LLM can act, we also have to control exactly what it can do, with which data and under what conditions. Key points: How the security model changes when an LLM goes from generating text to using APIs, tools, files and external services.; Understand the difference between direct prompt injection and indirect prompt injection and how external content can influence an agent's decisions.; Identify where trust boundaries appear when agents use MCP servers, tools, resources and APIs.; Limit permissions, credentials, tools and actions to reduce the impact of a compromised or manipulated agent.; Apply validation, allowlists, approval gates, logging, sandboxing and human-in-the-loop for sensitive operations.
AI/MLSecurity
Angel Jose
Cybersecurity & AI Security
2:40 PM — 3:10 PMLaboratorio
Workshops
¿Y si el problema de seguridad de tu aplicación comienza antes de escribir una sola línea de código? En esta sesión vamos a cambiar la forma en que pensamos el desarrollo de software: desde la definición del requerimiento hasta el código que finalmente llega a producción. Aprenderemos a identificar riesgos desde el inicio y a pensar como un atacante para descubrir cómo podría romperse una funcionalidad antes de que alguien más lo haga. A través de ejemplos prácticos exploraremos vulnerabilidades como SQL Injection, XSS, problemas de autenticación y autorización, validación de entradas y seguridad en APIs. Veremos no solo cómo ocurren, sino también cómo podemos detectarlas y prevenirlas durante el desarrollo. Porque un requerimiento también puede tener una vulnerabilidad, y una línea de código puede convertirse en una puerta de entrada. La pregunta es: ¿la estamos viendo antes de que alguien más la encuentre?
SecurityWebDevOps
Inés María Montero Concepción
Ingeniera en Sistemas | Full Stack Web Developer | Banking & Payment Systems
3:15 PM
3:15 PM — 3:55 PMPlenary
AI/ML4:00 PM
⏸
CLOSING
PLAN YOUR DAY
Take the schedule with you.
Subscribe and keep your calendar up to date. Or download a copy that never updates.
My schedule
No talks savedStar the talks you care about and build your own calendar.
Works with Google Calendar, Apple Calendar and Outlook
Timezone-aware (ADT / UTC-4)