I build AI agents and practical automation, and teach applied AI engineering in Spanish.
- What I do: Co-Founder and Technical Lead at MART Automations, where we build AI automations for companies in Panama. RAG systems, agents, and pipelines that have to survive real data.
- What I'm studying: Systems Engineering.
- What I'm into: the unglamorous half of AI engineering. Evals, failure modes, cost per run, and what happens on day two.
- Languages: Spanish (native) · English (B2) · French (B1)
I teach the engineering ideas behind useful AI systems, then give people a small exercise they can run and adapt for their own study. Linked talks have public resources for participants.
| When | What | Where |
|---|---|---|
| Oct 2026 | AI agents with Pydantic AI | PyCon Panamá 2026 |
| Oct 2026 | A system, not a conversation | CONTECS, FISC, UTP |
| Oct 2026 | Applied AI, not AI hype, with Álvaro Monagas for MART Automations | CONTECS, FISC, UTP |
| Oct 2026 | Judge, AI Agents Hackathon | FISC, UTP |
| Sep 2026 | Build AI agents with Pydantic AI | FISC AI Agents Hackathon, UTP |
| Aug 2026 | The AI fundamentals that do not expire | IEEE InnovaTech Congress, Universidad Interamericana de Panamá |
| Jul 2026 | From chatbot to personal assistant | FISC, UTP — delegation from Universidad Peruana de Ciencias Aplicadas |
| Jun 2026 | Build your first agent with LangChain | EurusConf 2026, Python Edition |
| Jun 2026 | Prompting and AI agent design | Universidad Tecnológica de Panamá |
At PyCon Panamá I showed how typed agents, FastAPI, and deterministic tests make an agent maintainable. At CONTECS, each participant set up their own AI operating system, wrote a skill, connected an external tool, and published a page from it. The FISC workshop goes from tools, MCP, skills, and memory to a working Pydantic AI agent. InnovaTech connects context and grounding to an AI workspace. The UTP talk for the Peruvian delegation shows how an academic workflow grows from a chatbot into an assistant. EurusConf takes a LangChain agent from Colab toward a local project. The earlier UTP workshop covers prompt design and the first agent loop.
The public repositories contain participant resources. My slides and speaker notes are separate. Each workshop repository explains its usage terms; permissions already granted for older CC BY 4.0 versions remain in effect for those versions.
- rtovar-aios — a template for students who want to run their own AI operating system. MIT. Used in the CONTECS workshop.
- Proposal agent (MART Automations) — internal agent that reviews the commercial context of an opportunity and prepares a diagnosis, technical scope, price estimate, and branded PDF proposal. The team reviews it in Slack before anything reaches a client. In pilot; private.
- Business operations agent (MART Automations) — agent for inventory, catalog, quotes, purchase orders, and Word and PDF reports, with modular skills, monitoring, integrity checks, automated tests, and human approval. Python, Linux, Docker. Private.
- AI editorial system for the Caja de Seguro Social de Panamá — selected among 360 applicants, I designed and deployed it to production in four hours. It classifies topics, recommends audience, channel, and format, and drafts pieces for human sign-off. Owned by the CSS.
