Welcome to HiddenLayer | Hidden Layer

Welcome to HiddenLayer

May 1, 2025 · 8 min read · introduction ai engineer career ·

Welcome to HiddenLayer

This is HiddenLayer — a space where I share, with total clarity and honesty, what it really means to work with Artificial Intelligence in 2025. Here, you won’t find LinkedIn buzzwords or empty hype. Instead, you’ll find:

  1. My journey as an AI engineer
  2. Behind-the-scenes of real-world projects
  3. Insane bugs
  4. Tough architectural decisions
  5. Lessons learned through hands-on experience

All of this from someone who’s spent years as a Data Scientist and chose to go deeper into the stack.

1. Who I Am

My name is Alex Rodrigues. I’m an AI engineer focused on Gen AI solutions, with over 10 years of experience in the data field. Before working with LLMs and production-grade neural pipelines, I held roles as an analyst, senior data scientist, and tech lead in both Brazilian and U.S. companies. This journey gave me a clear view of how real AI systems are built — far from playgrounds, and deep within business pain points.

In recent years, I’ve fully immersed myself in systems built with Retrieval-Augmented Generation (RAG), LangChain deployments, vector database integration, and everything it takes to turn research into real products. I currently work with major corporations and U.S. startups, delivering solutions that blend architecture, engineering, and AI-powered UX. I also hold a postgraduate degree in Artificial Intelligence and remain a constant learner — testing, iterating, and applying every day.

This blog was born from a desire to document and share that journey. No fluff, no buzz. Just real content, straight from the technical trenches. This is engineering — not decoration.

2. What is the HiddenLayer Project?

HiddenLayer is more than a blog. It’s a personal and technical logbook of what really goes on behind the scenes when building AI solutions in the real world. Here, I share my experiences as an AI Engineer — from architectural decisions and working models (and failed ones), to reflections on career, learning, and the gritty behind-the-scenes work no one talks about. No generic posts — everything here is battle-tested.

The name is symbolic. In neural networks, the “hidden layer” is where the deep processing happens — invisible to those who only see inputs and outputs. And in a tech career, it’s the same: there’s a lot of noise and showmanship, but very few people reveal what lies beneath. This project is my attempt to expose that hidden layer — with transparency and practical value for those who want to grow in the field.

3. What You’ll Find Here

This blog brings together three things: my hands-on experience as an AI engineer, my desire to build real technical content, and the lack of honest references about what it means to actually do this for a living today. You’ll find code, lessons, failures, difficult decisions, career reflections, paper summaries, LLM experiments, and architectural insights. No marketing filter. Just what works — and what fails in useful ways.

3.1 My Transition to AI Engineering

I spent over a decade working as a Data Scientist in large companies, covering areas like risk, product, and growth — delivering predictive models and insights with real business impact. But at some point, I realized that the data science role was stabilizing too much, while the heart of modern AI was shifting toward engineering. I wanted to be closer to the models, the architecture, the deployment. I wanted to build intelligent systems — not just analyses.

The transition wasn’t easy. Shifting from DS to AI Engineer meant leaving a comfortable analytical mindset and embracing a product-oriented engineering vision. I had to learn dataset versioning, inference pipeline control, latency handling, scalability, and tradeoffs between embeddings vs fine-tuning. It also meant relearning how to code — with a focus on engineering, not just notebooks — and accepting that many problems are infra problems, not model problems.

The main reason for this move? Simple: AI Engineer is the most strategic role in modern AI today. It bridges the gap between models and real users. That’s why technical interviews are tough — they demand architectural knowledge, modern NLP, clean code, testing, deployment, and clear reasoning. I had to relearn how to study. I had to relearn how to fail. Here, I’ll share everything that helped me: from scripts that saved me hours, to mental shortcuts that got me through interviews that once would’ve crushed me.

3.2 My Day-to-Day as an Applied AI Engineer

Today, I work as a contractor for U.S.-based companies, involved in projects ranging from building autonomous agents with LLMs to designing robust pipelines for document ingestion, chunking, vectorization, and retrieval. Some projects involve integrating models into critical systems, while others focus on enhancing internal user experiences through RAG and context-aware automation. Each one comes with its own architecture, stack, and flavor of chaos.

The hardest part isn’t the model itself — it’s the ecosystem around it: dealing with messy data, ambiguous contexts, and architectural decisions that directly impact latency and scalability. Managing vectors, embeddings, versioning, MLflow experiments, deciding between fine-tuning and prompt engineering — all while writing clean, testable, production-ready code.

And of course: the bugs. The errors that only show up in production. Pipelines that break on a poorly formatted PDF. Models that benchmark well but crash when given unexpected inputs. Real-world AI is far less about “training a model” and far more about making sure everything around it doesn’t fall apart. And that’s exactly the kind of story I aim to tell here — no sugarcoating.

3.3 Kaggle and Practical Competitions

Even though I work professionally with AI, I still participate in competitions — especially in areas like NLP, LLMs, and applied embedding-based systems. Kaggle is my testing ground. A place to try out ideas, validate hypotheses, and keep my learning mindset sharp. It’s also where I wrestle with new datasets, different constraints, and metrics that force you to break out of autopilot.

I’ll share my most relevant challenges here — with a focus on end-to-end solutions: how I approached the pipeline, the trade-offs I faced, and the code I wrote (including the mistakes I learned from). It’s all hands-on, no polish. The goal is to inspire those studying LLMs in a practical way — whether for interviews, projects, or career growth.

3.4 AI & LLM News That Actually Matter

AI is evolving at a dizzying pace. There’s a new paper every day, a promising library, a “revolutionary” model. But if you work with this stuff, you know: 80% is noise. Only a fraction actually changes how we build real-world AI. This section is my attempt to separate signal from noise.

You’ll find direct, critical takes on major papers, practical summaries, and commentary on what really matters in new releases. When something new comes out, I’ll test it, comment on it, apply it (when it makes sense), and give you a straight answer: “this is worth it” or “this is just buzz.” No fanboy hype. No tribalism. Just applied engineering with a critical lens.

4. My Plans for the Future

Though I’m deep in production projects, I maintain a growth mindset. I’ve been pursuing certifications in cloud, modern NLP, and LLM engineering — not for the acronym, but to organize my technical knowledge and unlock more strategic opportunities.

I’m also preparing for the next academic step: a master’s degree (and maybe a PhD later), with a focus on symbolic agents, neuro-symbolic AI, and intelligent system architectures. I want to deepen the theoretical foundations that support what I build every day.

And the blog? It will grow alongside me. I want it to become a practical knowledge base — but also a channel for exchange. Maybe a newsletter, some open-source code, public experiments. This space will evolve at the same pace I do. If you’d like to follow along, you’re more than welcome.

Thanks for Reading

If you’ve made it this far, thank you for your time and attention. I hope this space becomes useful to you — as a reference, inspiration, or a push to go further. The AI engineering journey is intense, full of uncertainty, but also full of amazing discoveries. That’s why it’s worth sharing.

If you want to exchange ideas, collaborate, or just follow my work, find me on LinkedIn or check out my repositories on GitHub.

Let’s enhance the future of AI together.