I.A New Chapter

In 2026, I worked across AI integrations with Salesforce, process automation, a fully AI-driven greenfield project, the large-scale adoption of Spec-Driven Development, and the refactoring of mission-critical Kafka systems that handle millions of messages a day.

II.Pioneering Spec-Driven Development

I was the first intern at Dell Technologies to adopt a high-efficiency AI development protocol, on a team pioneering AI-driven engineering. I validated new approaches, shipped at a high cadence, and helped teammates use the method daily.

III.AI Integration

In my first two months, I delivered an interface that lets business stakeholders edit the classification rules behind our AI and compare, on real project records, current results against the proposed rules.

IV.PowerShell Automation

After a reorganization, I joined a multi-micro-frontend project where running each app locally took several tabs, file edits and five commands per change. I built a PowerShell automation that does it in one command, with flags for each scenario, saving every developer 30+ minutes a day.

V.Expanding AI Integrations

I delivered integrations with external AI services and led an event-driven migration from a legacy service to a new AI tool, using feature flags and fallbacks so both versions coexisted for a gradual rollout.

VI.Building from the Ground Up

I built a new micro-frontend from scratch in three days with a fully AI-driven workflow, generating 10,000+ lines of configuration aligned with Dell.com architectural standards by mapping our other apps' classes and libraries into prototype classes.

VII.End-to-End Ownership

From the prototype, I took end-to-end ownership of a filtering feature, abstracting 50+ dataset attributes into a single type-aware filter model across backend and frontend, with custom drawers that meet accessibility guidelines.

VIII.Building Queues with RabbitMQ

Records with unresolved references wait for a product with a matching reference. I built a RabbitMQ queue that publishes an event for each new product, finds the records sharing its reference, and revalidates them automatically.

IX.Safeguarding Critical Systems

I refactored high-criticality systems to flag early which records can be processed independently. Those without ordering requirements now run through a dedicated batch processor while the rest keep their ordered path, removing a major bottleneck and stabilizing peak hours.

I.Program Overview

I was accepted into Dell's IT Academy after applying alongside more than 700 candidates: a 6-month, immersive program covering full stack development, AI/ML fundamentals, and databases, before placement onto a team. Beyond coursework, my cohort built practical systems across .NET MVC, Spring Boot, Angular, and Jupyter Notebook.

II.Month 1: Foundations

We started with versioning, OOP, and SQL, then moved into C#, ASP.NET, Entity Framework, and SQL Server. The practical exercise: running mock client sessions to gather requirements and build a working system with .NET MVC, Razor Pages, and SQL Server.

III.Month 2: Frontend & DevOps

This month covered HTML, CSS, TypeScript, and Angular, along with early DevOps work on an AWS instance. The practical exercise: an IT resource allocation system built with .NET, Angular, and PostgreSQL.

IV.Month 3: Java & Spring

We picked up Java, Spring Boot, Spring MVC, Spring Data JPA, and MongoDB. The practical exercise: a contract management system for starting projects, setting salaries, and calculating project costs by duration.

V.Month 4: Machine Learning

The program pivoted to ML: Linux CLI, Python, NumPy, Pandas, ETL concepts, Scikit-learn, deep learning, and LLMs. The practical exercise was a full case study on a Kaggle dataset of our choosing. We picked one that predicts user emotion from a photo.

VI.Months 5–6: Capstone Project

A simulated internal system, inherited and extended across 6–8 prior cohorts of the program, managing an internship program end to end with matchmaking logic between candidates and open positions. We ran it with full Agile ceremonies: a Product Owner, sprint planning, and burndown tracking.

VII.My Contribution

Our cohort added a team-rotation process for internal employees. My part:

• ·Built the entity model and database from scratch
• ·Designed the entry points into our WebSocket interfaces
• ·Added data pages, bug fixes, and UX polish

VIII.Next Steps

After finishing our contributions on the Capstone Project, we were assigned to a designated team in Dell Technologies.

I.Context

After getting a perfect grade on six Probability and Statistics assignments, I was invited to work as a teaching assistant for a semester, before my evening classes. The subject is mostly tied to Computer Science and Engineering, but it's taught across many courses at PUCRS, including Business, Psychology, and Journalism.

II.Working With Students

Over those months I worked directly with a wide range of students, adapting sessions for accessibility needs and different levels of comprehension, including students with autism spectrum disorder (ASD). Several students set aside a recurring time each week just to come by and go over that week's material with me.

III.Results

We got strong results over the semester. One student I tutored came in with a 4.5 on the partial exam and closed the semester with an 8.0 final. I built a method for this: I'd create a template using the simplest problem on a list, then apply it repeatedly until students picked up the pattern on their own, working independently rather than depending on me.

IV.A Pattern Worth Studying

By the end of the semester I put together a small case study based on what I noticed day to day. Most students struggled with certain exercises simply because they avoided scientific calculators, sticking with older, simpler models even though the coursework required the more advanced ones, along with basic concepts of how to structure equations on that type of calculator.

V.Feedback That Mattered

This showed up constantly: more than 80% of the questions I got weren't about the material itself, they were about how to operate a scientific calculator (order of operations, how to write out full formulas). I passed that feedback to my supervisor to help build a short workshop on scientific calculator basics.

VI.Publication

Estudo sobre a Aplicação de Visualizações Narrativas em Dados Governamentais (2025).

I.Context

For 14 months, I researched under Prof. Dra. Eliane Schlemmer at GPedU. The work moved between virtual-world fieldwork, national-scale qualitative research, teaching support, and public communications for the group, and no two weeks looked quite the same.

II.Metaverse Research

I gave live technical support during research interviews conducted inside Second Life, then helped design and populate a new environment in Mozilla Hubs, starting from a participant questionnaire and ending with a space people could actually walk through. I also learned Zappar and Spark AR well enough to prototype a public augmented reality demo.

III.Qualitative Analysis at Federal Scale

I transcribed and coded live meetings between Brazil's State Secretariats of Education and UNDIME, tracking how each region of the country talked about post-pandemic education. I started in NVIVO and moved to MAXQDA as the project grew — my first real exposure to qualitative research methods.

IV.Postgraduate Teaching Support

I supported a postgraduate class, Ensino e Aprendizagem no Mundo Digital, helping students navigate unfamiliar digital tools. Separately, I worked with Applied Linguistics master's students building a custom metaverse for their own research.

V.Research Impact Mapping

I used Google Earth to trace the group's global reach: livestream locations, citation trails from its most-read publications, and an interactive map following the onLIrE international education network from São Leopoldo to Sweden.

VI.Digital Presence & Production

I maintained the group's WordPress site, updating publication records and fixing broken links. I designed cover art for GPedU's international webinars, edited event video, learned basic StreamYard operation for live broadcast support, and presented during the closing session of We, Learning with the Cibricity (WLC).

VII.Organizing the BVER

My final project built on a master's student's research: Rodrigo Lehnemann's Biblioteca Viva Ecológica Reticular, a platform that organizes collective memory in a reticular, graph-based structure. I populated it with data from the group's teacher-training initiatives and built the accompanying Mapa da Formação in Miro, turning years of program activity into something a stranger could actually navigate.

VIII.Presenting the Work

I presented the finished project at UNISINOS' XXV Mostra Científica, alongside a full presentation on GPedU's research delivered to an academic audience — the first time I had to explain the group's work, from the ground up, to people outside it.

IX.Publication

EscutAÇÕES, ConversAÇÕES e CompartilhAÇÕES enquanto práticas conectivas na formação de professores-pesquisadores (2023).

I.Origin

In late 2021 I started designing and publishing augmented reality filters for Instagram. Within two years my effects had been used in more than 1 million Instagram Stories worldwide and reached over 80 million impressions.

II.Platform Context

Spark AR was Meta's platform for building and publishing Instagram filters. Creators could publish effects on their own profiles or take custom commissions from brands and other users. Getting there meant teaching myself photography, programming, and 3D modeling, and turning those skills into a product people actually used.

III.Anatomy of a Filter

A filter that actually works needs several pieces:

• ·A color LUT, converting the camera's raw feed into a defined preset look
• ·Custom textures, built from scratch
• ·An interactive interface, designed to keep people engaged
• ·Custom face-sculpting logic

IV.Scaling It Like a Product

Getting a filter to spread wasn't just a technical problem:

• ·Reading platform KPIs to understand what was actually working
• ·Building a marketing plan for a global, not local, audience
• ·A working knowledge of code. Spark AR could technically be built no-code, but the node wiring gets genuinely hard to manage at scale

V.Growth & Iteration

Every launch taught me something about running a small product. I ran constant A/B tests on interface details, watched how small changes moved user behavior, and used those trends to decide what to build next. The filters reached millions of accounts through iteration, not luck.

VI.Freelance Work

As the account grew, local businesses started reaching out directly. I built a custom filter for a merger between two companies in Rio Grande do Sul. The one I'm proudest of is a blood donation awareness filter, where users tapped through a custom interface to select and share their blood type as a call to action.

VII.The End of an Era

Meta announced in 2024 that it would retire the Spark AR ecosystem. By 2025 every filter I'd published was gone, along with all my metrics. I shifted focus after that, toward landing a software engineering internship.

Figures reflect the last dashboard snapshot available before the platform shut down.