← Founder page

Podcast and speaking

I bring three angles to the conversation.

Founder, AI systems builder, Latina based in Bogota. I open-sourced the operating system I built to run two companies. Hosts can validate the work in 30 seconds before the call.

Shark Tank at 18 (Cuban + Corcoran) · 2 exits across 7 ventures · Success Magazine 30 Under 30

Featured 40+ times in

TODAY Show · Forbes · Inc. · Entrepreneur · Business Insider · Authority Magazine · Thrive Global

Guest lectures at

Columbia · Wharton · NYU Stern · SMU Cox · General Assembly

Three angles. Pick the one that fits your audience.

Don't send "Shark Tank" to a tech podcast. Don't send "open-sourced my AI brain" to a LatAm founder show. The angles are pre-shaped to specific listener types so the booking conversation can skip straight to scheduling.

Angle 01

I open-sourced my company's AI brain

For AI and tech podcasts, dev tool shows, the Claude/Anthropic ecosystem, Obsidian and PKM communities, productivity shows for operators.

I built an AI operating system to run two companies from Bogota, then open-sourced the core. The visible piece is a public concierge at concierge.diazroa.com that recognizes guests by name, routes nine intent types, and files reasoning briefs back to my vault. Underneath: Claude Code plus Obsidian, 9,000+ knowledge nodes, custom skills, advisory panels, pattern recognition from my own journals.

The episode can land on:

  • Why "one brain" beats "fifteen tools"
  • The specific stack: Claude Code, Obsidian, MCP servers
  • What happens when you give your journal to an AI and ask it to find patterns
  • The case for open-sourcing your own operating system, even if you sell consulting on top
  • What consulting looks like when the product is installable in a week
  • The Concierge: public-facing AI that holds your voice without breaking the spell

Useful for prep: talk to the live Concierge · browse the open-source core

Angle 02

From Shark Tank at 18 to building AI systems

For founder podcasts, startup shows, "journey" episodes, 30 Under 30 adjacent shows, women founder shows, entrepreneurship podcasts.

I was 18 when Cuban and Corcoran invested. Since then I've built seven companies, had two exits, and spent the last year building an AI operating system for the ones I'm running now. What I'd tell my 18-year-old self isn't about the TV deal. It's about the thing I'm building a decade later: the system matters more than the idea.

The episode can land on:

  • What Shark Tank actually did, and didn't do, for the business
  • Seven companies, two exits. What the other five taught me.
  • The move from scrappy consumer products to B2B infrastructure
  • Why most founders underinvest in their own operating system
  • What I'd build first if I were starting today

Useful for prep: the consulting practice that emerged · the deep-dive essay

Angle 03

Why LatAm founders think differently about AI

For international and LatAm shows, emerging-markets tech, Colombia and Mexico-focused podcasts, Spanish-language shows, global founder shows.

I run a company from Bogota that's raising a US seed round. Most of my peers are building with constraints US founders never had to think about: cheaper talent, smaller domestic markets, distribution that's mostly WhatsApp and in-person, clients who expect white-glove service at SaaS pricing. AI changes the math, but not the way most US tech podcasts describe it. The LatAm version is different.

The episode can land on:

  • Why LatAm corporate sales is so different (relationship density, trust cycles)
  • How AI changes unit economics for small teams in emerging markets
  • What US investors miss about LatAm market size (hint: it's not a smaller US)
  • Building bilingual AI systems (English and Spanish, not translation)
  • Why outsourcing to LatAm is 2019 thinking; building AI-native companies in LatAm is the next wave

Useful for prep: the Spanish-language essays at perspectivasblog · talk to the Concierge in Spanish

Built. Open-sourced. In production.

Before you book me, talk to the system. The Concierge runs my real inbound. The repo is what it's built on top of. The Substack post is a written deep-dive in my voice.

The one-line

Talk to the system before you talk to the speaker.

The Concierge runs my real inbound. Spend thirty seconds with it and you'll know if I fit your show. Open-sourced core, bilingual, eight months in production across two companies.

Specific stories, not just topics.

Each one is a real moment, not a hypothetical. Hosts can pick three and we shape the episode arc around them on the call.

The Concierge that doesn't break the spell

Recognizing a guest by name without announcing it, the Jackie Kennedy executive-assistant pattern translated into AI. Why "are you X?" is the wrong move and what to do instead.

When my AI caught me about to repeat a pricing strategy I'd already dropped

Pattern recognition from journals. The system spotted a six-month-old decision and surfaced it before I made the same mistake.

Seven companies, two exits, what the other five taught me

Pawliday Inn (sold), Robin House (running), Onde (raising), and four that closed. The lesson isn't survivorship.

Open-sourcing the operating system you sell consulting on top of

Counterintuitive move that pulled qualified buyers in faster than any cold outreach. Why the moat is the install, not the code.

Bilingual AI for bilingual lives

Building knowledge graphs and skills that work in English and Spanish without translation drift. What changes when your second brain is bicultural.

The advisory panel inside the system

Thirty named voices weigh in before any high-stakes decision. What it means to take dissent seriously when the dissenter is an AI prompt.

Solo, panel, or workshop. Pick the format.

Format 01

Podcast guest

45–90 minutes

Long-form conversation with one host. Audio or video. I bring the angles, hosts shape the arc. Recording in the show's preferred platform (Riverside, Zencastr, Zoom, in-person if Bogota or NYC).

Format 02

Live panel or fireside

30–60 minutes

Conferences, summits, member events. Comfortable on a panel of 3–5 or as a single fireside guest with a moderator. Bilingual (EN/ES) for LatAm and bilingual events.

Format 03

Workshop or talk

45 minutes – 2 hours

For founder communities, accelerators, AI-curious teams. "How I built my AI operating system" or "What an AI EA actually looks like in production." Live demo of the Concierge, Q&A, takeaway one-pager.