Miguel Amigot

At a Banco Santander gathering of roughly 100 university presidents in Mexico City, Miguel joined a panel — moderated by Paul LeBlanc — on what agentic AI makes possible in higher education today, and what is still hard.

The throughline was AI sovereignty and ownership. Institutions don't have to lock themselves into a single vendor's models at $20–$30 per user per month. By staying LLM-agnostic and tapping developer-tier token pricing, a deployment that would cost millions a year can run for a fraction of that.

Miguel made the case for owning both the code and the data — running the entire platform inside the university's own infrastructure — so security, cost, and the roadmap stay under the institution's control rather than a SaaS vendor's.

A Google Cloud session on responsibly driving student success with AI — how institutions deploy AI for measurable student outcomes while keeping their data governed and under their own control.

In CUNY's AI Roundtable Discussion Series — “Challenges of AI in Higher Education” — Miguel sat down with Jose Diaz of CUNY City Tech to push past the hype and into the strategic realities institutions can't ignore.

The core idea: universities don't have to lock themselves into expensive, closed AI ecosystems. They can build LLM-agnostic infrastructure that mixes and matches models — OpenAI, Anthropic, Google, open-source, or local — while keeping control of their data and costs.

The conversation covered real use cases: scaling advising, tutoring, admissions, and financial-aid support; reducing operational cost while expanding access; and improving equity through always-available, personalized student support — all with institutional data governed internally.

It also confronted the harder questions — what happens to teaching and learning under AI, how to handle academic integrity and cheating, the evolving role of faculty (especially in the humanities), and how to balance personalization with critical thinking.

The takeaway: AI adoption isn't optional, but how you adopt it is a strategic choice. Flexibility, cost-awareness, and data control separate the institutions that lead from those that get locked into short-term decisions.

At SUNY's Conference on Instruction & Technology (CIT 2026) at Stony Brook University, Miguel walked through the journey universities take from free chatbots to full code and data sovereignty.

The talk traces seven stages — from free ChatGPT accounts with no governance, through enterprise agreements and the multi-vendor cost problem, to the developer-pricing revelation where the same models cost ~100× less via API.

From there it builds toward LLM agnosticism, an integration layer that plugs into the LMS and SIS, and finally code ownership — the institution running and owning the entire stack, with private models running on its own infrastructure.

The conclusion for SUNY: institution-owned AI isn't just cheaper, it's the only model that keeps data, governance, and the long-term roadmap fully in the institution's hands.

A conversation with Amazon Web Services on bringing AI to education — building agentic AI on cloud infrastructure that institutions can run, govern, and own end to end.

An engineering talk for the IEEE high-performance-computing community, hosted at MIT Lincoln Laboratory: the architecture behind production AI mentors — custom retrieval indexes, prompt design, guardrails, and the APIs that tie them together.