Always There, Always Aware, Always Connected to Your Real Mentor

March 1, 2026
Luis Hernández

Category: AI in Education, BTLPT, EdTech, Professional Development, Teacher Education

AI-Generated Audio Summary

When I started my first year as a bilingual teacher, I had plenty of theory but very little idea of what to do when twenty-five students were staring back at me, each with different levels of English proficiency, different backgrounds, and different ways of processing the world. I was overwhelmed.

What saved me were mentors. Experienced teachers who took me under their wing and helped me navigate a system I did not fully understand. I had arrived from Venezuela with strong content knowledge and a deep belief in education, but the American school system was new to me. The culture was different, the expectations were different, and the rules (both written and unwritten) were different. Those veteran educators helped me make sense of it all. They shared what worked in their classrooms. They helped me bridge the gap between what I had studied and what was actually happening when I stood in front of my new multilingual students. That mentorship made me the educator I became in this country.

But even with their support, I still struggled. There were challenges that mentorship alone could not fully address, particularly when it came to reaching every student in a bilingual classroom with different language needs, different backgrounds, and different ways of processing content. I started looking for other forms of support, and I found one in technology.

Finding a Second Mentor in Technology

I have always believed that education and technology belong together, not because technology is impressive, but because it can extend what good teaching already does. During my years in the classroom, I completed a master’s degree in Computer Education and Cognitive Systems at the University of North Texas, where I studied coding, computing, and how people learn with and through technology. That program gave me the language and the tools to connect my two passions: education and technology.

One of the earliest inspirations for this connection came from a place much closer to home than a university. My grandfather, who raised me alongside my mother, taught me the value of creativity, observation, and thinking deeply about the world. He was not a technologist or a scientist. But he showed me, through the way he lived, that paying attention to things, asking questions, and making connections was its own kind of intelligence.

Years later, when I encountered Leonardo da Vinci’s notebooks, everything my grandfather had cultivated in me clicked into place. Here was someone who had turned observation, curiosity, and the connection between art and science into a way of understanding the world. Not the artist most people think of, but the scientist, the systems thinker, the engineer, the person who wrote that to develop a Complete Mind, you must “study the science of art, study the art of science, develop your senses, and realize that everything connects to everything else.” I was no Leonardo, of course. But I recognized something in that philosophy that felt familiar, something my grandfather had planted in me long before I had words for it. And I realized I could build on that same foundation: connecting creativity with technology, observation with design, and art with learning.

I started creating resources to help bilingual teachers teach science and social studies, with a particular focus on developing academic vocabulary through visual literacy, video, animation, and interactive strategies. Over the years, those resources were acquired by a publisher in Texas and distributed as a series of books and companion DVDs. Yes, DVDs. That will tell you how long I have been at this.

But here is the part that matters. Whenever the publisher invited me to present, I had one request: take me to the school districts and put me in front of the teachers. Because the DVDs were never the point. The strategies behind them were the point. How a teacher uses a tool matters infinitely more than the tool itself, that was true with video and it was true with interactive notebooks. And it is true today with artificial intelligence.

I wrote about this philosophy, the journey from constructivism to constructionism, and how it has shaped my approach as both a teacher and an instructional designer. The core idea has never changed: technology is building material for learning, not a shortcut around it.

The Challenges That Still Exist

The problem is that, still today, not every new teacher gets the kind of mentorship I was fortunate to receive. And the challenges I faced when I first entered the classroom have not gone away.

After years of teaching in bilingual classrooms, I moved into higher education and worked as a professor at a couple of universities here in Texas, preparing the next generation of bilingual and ESL teachers. That experience gave me a different vantage point. I saw firsthand, from the other side of the table, just how hard it is to prepare teacher candidates for the realities they will face. Teacher preparation programs are doing important work, but they face real structural challenges that are difficult to solve with traditional approaches alone:

Balancing institutional and student needs. Many teacher candidates juggle jobs and family responsibilities, making it hard for programs to structure training that works for everyone.
Bridging theory and practice. Many programs do excellent work with clinical experiences and school district partnerships, but not every candidate has equal access to those opportunities. Even when they do, the range of scenarios a student teacher encounters during a placement is inevitably limited.
Supporting academic language development. Many bilingual preservice teachers struggle with mastering subject-specific vocabulary and professional communication in both Spanish and English, a challenge that preparation programs have limited time and resources to address individually.
Meeting evolving certification standards. Here in Texas, for example, the new TExES 165 Bilingual Education Spanish Supplemental exam launching in September 2026 merges what were previously two separate exams into one integrated assessment. Programs must now prepare candidates to demonstrate pedagogical knowledge through academic Spanish, not alongside it.

These are not failures of any program. They are structural realities that even the most well-designed programs face. And this is exactly where I believe AI, built the right way, can help.

Why I Started Building AI Tutors

For years, I had been creating technology-based educational products, always with the same goal: put good pedagogy into the hands of teachers and learners. The tools changed, from printed materials to DVDs to web applications, but the approach stayed the same. Structure and strategy first, as well as making sure that technology is always in service of learning.

Then, about four years ago, something shifted. Large language models arrived, APIs became accessible and generative AI made it possible to build systems that could hold natural conversations, respond to context, and adapt in real time. I saw the perfect combination of elements.

Everything I had spent years developing, my knowledge of education, my training in computer science, my experience coding with HTML, JavaScript, and PHP, my understanding of how bilingual learners acquire language, it all converged. For the first time, I could build what I had always wanted to build: AI tutors that function like the mentors who had made such a difference in my own career. Not chatbots that answer questions, but AI tutors that meet learners where they are, scaffold their growth, and provide the kind of guided support that a skilled mentor provides, available whenever the learner needs it, grounded in real expertise.

I was thinking about the bilingual teacher candidate working a full-time job who cannot meet with a tutor during office hours. The preservice teacher in a rural area without access to a bilingual education specialist. The student who needs to hear the same concept explained three different ways before it clicks.

I have written about this journey many times on this blog, from why educators and technologists need each other in the age of AI to how AI is helping us redefine language development to how we are using AI to redefine teacher preparation. The core belief has never changed: technology serves pedagogy, not the other way around.

AI Tutors Grounded in Vygotsky

Lev Vygotsky’s concept of the Zone of Proximal Development, the space between what a learner can do alone and what they can do with guided support, has always been central to good teaching. A skilled mentor meets you where you are and helps you reach where you need to be.

This is the exact framework we use to design our AI systems. We are not building chatbots that answer questions. We are building AI tutors that function as Vygotskian mentors: context-aware, knowledge-grounded, and designed to scaffold learning at the level the student needs.

Here is what that looks like in practice:

Expert-curated knowledge bases. Our AI systems do not rely on generic training data. They are built on vector databases filled with curated content from bilingual education specialists, certification competencies, classroom scenarios, and instructional frameworks. Every response is grounded in verified, domain-specific knowledge.
Structured learning interactions. Learners do not just chat freely with the AI. They interact through structured prompts designed for specific domains: reading, writing, listening, speaking, each processed through pre-engineered evaluation criteria.
Adaptive, research-backed feedback. The system provides actionable guidance, not just scores. It tells learners what to improve and how, using frameworks like the ADG model (Acknowledge, Describe, Guide).

We introduced our AI tutor, Veronica, over a year ago. Since then, she has evolved significantly. She now draws from a knowledge base that includes bilingual education competencies, special education frameworks, BTLPT certification content, and curated expertise from specialists in the field. She is context-aware, meaning she adapts her responses based on what the learner is studying, and she is designed with strict anti-hallucination guardrails to ensure accuracy.

Toward Dual-Axis Evaluation

Here is where things get interesting. For the past year, we have been working on a project that goes beyond conversational tutoring.

The real challenge in bilingual teacher preparation is that candidates need to develop along two dimensions simultaneously: they need to master pedagogical knowledge (classroom management, bilingual education theory, assessment, differentiation) and they need to demonstrate target-language proficiency (academic vocabulary, grammatical accuracy, professional register, fluency). Traditional assessments measure these separately. But in the real world, and in the new TExES 165 exam, they are inseparable.

We have built an AI system that evaluates both dimensions at the same time, in context. The AI presents a realistic professional scenario, the learner responds in the target language, and the system evaluates both the quality of their pedagogical reasoning and the quality of their language use, and provides specific, actionable feedback on both.

This is what we are calling dual-axis evaluation, and we believe it represents a meaningful step forward in how AI can support multilingual professional preparation.

The content pipeline behind this system is also worth noting. We use a model of expert-AI collaboration: domain experts create validated master scenarios from authentic professional situations. AI generates variations across complexity levels and language demands. Experts then review and approve the AI-generated content. The result is a validated, adaptive practice library, expert-quality content at AI speed, with human oversight at every stage.

But the AI tutor is only one side of the system. On the other side is the professor. Every interaction a student has with the AI generates structured insights that the professor can review: where the student struggled, what patterns emerged in their language use, and where they need human guidance. The AI does not replace the mentor. It extends the mentor’s reach and creates a window into each student’s learning process that would be difficult to achieve through traditional methods alone. When the student needs support that only a human can provide, the connection is already there.

We documented this work in a co-authored chapter with Dr. Uvaldina Janecek, published by IGI Global: “AI-Guided Professional Development: A Vygotskian Approach for Teachers in Multilingual Classrooms” in the volume Harnessing AI’s Potential to Support Student Success and Teaching Excellence.

What We Will Be Sharing at Texas State

On Thursday, March 5, 2026, Dr. Uvaldina Janecek and I will be co-presenting at the Texas State University AI & Teaching/Learning Symposium in San Marcos, Texas. Our session is titled “AI Tutors as Vygotskian Mentors: Scenario-Based Learning, Adaptive Feedback, and Target-Language Development.”

The session will cover:

Why the mentorship model matters for AI design in education
How we build AI systems grounded in expert knowledge and pedagogical research
The concept of dual-axis evaluation for multilingual professional contexts
A live demonstration of how these systems work in practice

As part of the session, we created an interactive visual walkthrough of the RAG architecture behind our AI tutors, covering how embeddings work, the full retrieval pipeline, the four layers of safety guardrails we use to prevent hallucination, and how all of it connects back to Vygotsky’s Zone of Proximal Development. If you are curious about the technical design behind these systems, it is worth exploring.

This is not a talk about AI in the abstract. It is about real systems, built by educators, grounded in research, and already being used with real learners.

We will also be sharing ideas about how institutions of higher education and teacher preparation programs can think about integrating these approaches into their own work. This is not about replacing what programs are already doing. It is about providing additional support, extending the reach of their faculty, and giving their students access to tools that adapt to their individual needs.

An Invitation

If you are attending the TXST AI & Teaching/Learning Symposium, I would love to see you in Room 315 at 2:30 PM on March 5. Whether you are a faculty member, an instructional designer, an administrator, or a teacher educator, I think this conversation will be worth your time.

And if you cannot make it to San Marcos, I encourage you to explore some of the ideas I have written about over the past few years. Here are a few posts that connect directly to what we will be presenting:

The question I keep coming back to is simple: How do we give every learner access to the kind of guidance that the best mentors provide? AI tutors will not solve that question on their own. But built with care, grounded in research, and guided by educators who have been in the classroom themselves, they can help us get closer.

I hope to see you there.

How We Use AI, and Why Our Digital Credentials Require Evidence

When we talk about our work with educators, at conferences and with universities, two questions come up: how do we use AI, and what does our digital credential actually prove? We answered both in writing, in two new pages on our website. The first is a note from our founder, Luis J. Hernández, on how an educator came to build with AI and where we draw the line. The second explains why our digital credentials are issued only when a learner shows the skill, and why that matters for learners and the institutions that prepare them.

A single objectives card pinned on an elementary classroom wall, with a stack of cards for the rest of the week on the teacher's desk.

One Objective on the Wall. Five Levels in the Room.

Principals want the ELPS posted, but one classroom holds students at five proficiency levels. How do you post one objective for all of them? We answered with something teachers can print and use, the Objectives Card in the ELPS Online Helper: one content objective, one language objective, and a strip showing what meeting it looks like at each level. Here is what it does, what we added after seeing how teachers used it, including a week planner and cards for lessons taught in Spanish, and why it matters in a dual-language classroom.

Layered translucent glass panes suspended in dark blue space, each etched with glowing abstract letterforms from a different writing system, overlapping and interlocking where the panes cross.

Verónica Speaks Twelve Languages. The Hard Part Was the Rules.

Verónica, the AI tutor inside our courses and apps, now responds in twelve languages. Arabic, Telugu, Tamil, and Korean joined because districts and teacher preparation faculty kept describing the families they serve. Adding the languages was never the hard part, since these models were multilingual from the start. The hard part was the rules: which characters even count as letters, whether the page or the person gets to decide, and when a tutor should hold the target language instead of following a student into English. Here is what broke, what we changed, and why it matters.

A Certificate No One Can Verify. It Is Just a File.

Most professional certificates share one weakness: confirming they are real means contacting the issuer and waiting, so in practice it rarely happens. A PDF can be edited in a browser. A paper certificate proves only that someone owned a printer. This post makes the case for digital credentials for both learners and institutions, looks at what the 2026 hiring research actually shows, explains why course completion is the wrong trigger for issuing one, and describes transparently how our scored capstone gates work, why we chose the Proof of Knowledge platform, and what we would tell any institution starting from zero.