I Built a Personal Chatbot, and You Should Too
Building a personal AI assistant is more than having your own ChatGPT. Here's what I learned, how proactive flows changed my daily routine, and why the building process itself is the real reward.
About me
I build generative AI products that move past the prototype and reach real users, and I take care of what keeps them reliable in production: automated testing, guardrails, and data privacy.
I'm currently on UOL's Innovation team, where I turn ideas into LLM-powered MVPs and take the ones that prove themselves all the way to production. Before that, I spent almost four years on UOL Ads as a full-stack engineer. That's where I built the foundation many AI engineers lack: running systems in production, working with microservices, and shipping products end to end, from the UI to the pipeline.
Highlights
- Architected UOL Ads' AI video ad generator as a white-label solution, in production since 2025.
- Structured production quality for a multi-product chatbot used across UOL brands and ingresso.com, serving 30,000+ sessions per month: automated response testing, guardrails (including prompt injection prevention), and personal data anonymization for LGPD (Brazil's data protection law) compliance.
- Implemented hybrid search (BM25 + semantic) with multilingual embeddings over a Portuguese corpus on self-hosted Elasticsearch.
- Led internal sessions on AI-assisted software development for around 150 people.
Stack: Python, FastAPI, LangChain, LangGraph, Elasticsearch, Redis, PostgreSQL, Next.js, AWS (Lambda, S3, Bedrock), and LLMs from OpenAI, Anthropic, Gemini, and DeepSeek.
I write about what I build on this blog, like Crode, the personal AI agent I built with LangGraph and multi-model orchestration.
Work experience
Software Engineer, AI & Innovation
UOL, Remote
May 2026 - Present
I explore, prototype, and ship generative AI products to production for UOL group brands.
• Structured quality engineering for a multi-product chatbot (LangGraph, RAG on Elasticsearch, Redis, Next.js) used by UOL Ads, ingresso.com, and other UOL products: automated response evaluation across system prompt variants, guardrails (including prompt injection prevention), and personal data anonymization for LGPD compliance. It serves 30,000+ sessions per month for 10,000+ users.
• Standardized the team's development workflow with uv and automated tests, bringing predictability to a fast-prototyping team.
• Implemented the chatbot's RAG pipeline on self-hosted Elasticsearch, with hybrid search (BM25 + semantic), multilingual embeddings for Portuguese content, and continuous ingestion of a constantly updated FAQ.
• Building an AI assistant for customer service over WhatsApp in customer negotiation flows.
• Deliver POCs and MVPs in short cycles with Python, FastAPI, and LangChain, across multiple LLM providers (OpenAI, Anthropic, Gemini, DeepSeek, Bedrock).
• Led sessions on AI-assisted development (specs, task breakdown, small PRs) for around 150 people.
Software Engineer, UOL Ads
UOL, Remote
Jul 2022 - May 2026
Full-stack engineer on UOL's advertising platform: frontend, backend, and microservices, while driving the product's first AI initiatives.
• Took part in R&D and architected UOL Ads' AI video ad generator as a white-label solution, in production since 2025, with 500+ videos generated so far.
• Led the frontend migration from Vue 2 to Vue 3, moving the build to Vite (20+ screens, 200+ components), ensuring compliance with Vue 2 end of life and making builds more than 10x faster.
• Built and maintained microservices in Java/Spring and Node.js and the Vue.js frontend, from production support to feature delivery.
• For about 3 years, balanced the product roadmap with AI R&D work, until taking on that front full time.
Trainee
Asa Branca Aerospace (UFPE)
Oct 2020 - Jan 2021
• Developed the tracking module for a CubeSat in C++ as part of the control team in UFPE's university extension project.
Education
Full Stack Web Development
Trybe
2021 - 2022
Bachelor's in Electrical Engineering
Universidade Federal de Pernambuco
2018 - 2021
Skills
The stack I use to build and ship AI products.
AI
LLMs, RAG, LangGraph, LangChain, AI Agents, LLM Evaluation, Hybrid Search, Vector Search / Embeddings, Prompt Engineering, AI Safety (guardrails, prompt injection prevention)
Backend & Data
Python, FastAPI, Elasticsearch, Redis, PostgreSQL, REST APIs, Microservices, Java/Spring, Node.js
Frontend
Next.js, React, Vue.js, TypeScript
Cloud & DevOps
AWS (Lambda, S3, Bedrock), Docker, GitHub Actions
Let's Talk
Open to AI Engineer roles (remote) and to conversations about taking LLM products, RAG, and AI agents to production.
Explore the articles of my blog
View all posts »
Introducing ScrapeGraphAI with DeepSeek
Learn how to use ScrapeGraphAI with DeepSeek to automate web scraping using LLMs. This guide covers setup, real-world examples, schema-based extraction, and performance logging.