São Luís, Maranhão, Brazil
Throughout 8+ years of experience, I specialize myself in designing and delivering solutions that improve efficiency, reduce operational costs, and unlock new product capabilities. My work sits at the intersection of AI, automation, and data engineering, enabling companies to move faster and make better decisions. 📊 Key Business Impact with my expertise: - Helped an Europe-based early-stage startup building an AI companion platform from 0 to 1 reaching 40k of active users in the platform. Still going here as the Principal Backend Engineer. - I've crafted from the ground up a customer support AI (RAG + LLM, Embedding, Vector DB) for a Brazil-based insurance company, delivering human-like assistance, improving scalability and operational efficiency in roughly 96–99%. - I've crafted from the ground up a high impact event driven system for back-office operation responsible for processing large data ingestion for a Brazil-based e-commerce creator company. The result was a platform that reduced onboarding timelines by 50% and gave the business much greater operational flexibility. - I've streamlined a Brazil-based company reducing manual effort in tax review processes by up to 90% by implementing a rule-learning system with human-in-the-loop, where analysts-reviewed cases were persisted and automatically processing ongoing and future documents with similar tax context. 🛠️ Technical expertise: - AI: Google Document AI, LLMs, RAGs, Evals, Embeddings, Vector Search, LLM observability - Automation: OCR, Crawling, Scrapping, Speech Recognition, Classification - Data Engineering: ETL, pipelines, orchestration, data warehouses, Airflow, Dagster, Jupyter Notebook - Backend: Typescript, Node.js, Nestjs, Python, FastAPI, Docker, Redis, MySQL, MongoDB, ClickHouse, PostgreSQL, RabbitMQ, NATS. - Infra: AWS, GCP, Heroku, Cloudflare, OpenAI, Google AI Studio, Groq, RunPod, ElevenLabs - Observability: Grafana, Prometheus, New Relic, Sentry, LangFuse 🧠 What sets me apart: - Strong product mindset: I build with business impact in mind - High autonomy: manager of one with consistent deliveries - Proven ability to thrive in early-stage startups and greenfield projects - Continuous learner: always adapting to new technologies and challenges - Experienced working with cross-functional, international teams, stakeholders and customers
Principal Backend Engineer driving technical strategy, product architecture and engineering decisions while building core platform capabilities across AI, payments, infrastructure, and real-time systems from 0 to 1. 🚀 Business Impact - Built the backend platform from the ground up, enabling the company to launch its MVP and rapidly iterate based on user feedback - Enabled product monetization by building subscription management, recurring billing, and multi-provider payment infrastructure - Scaled the platform to 40K+ active users - Reduced infrastructure costs through cache optimization, database query refactoring, and efficient AI inference architecture - Improved AI cost visibility by building an internal LLM observability platform tracking token usage, model costs, latency, and prompt performance - Enhanced platform safety by introducing automated AI-powered content moderation for text and image generation - Designed a scalable backend architecture capable of supporting conversational AI, image generation, real-time messaging, and future product expansion 🔧 Core Contributions - Partnered closely with founders to translate product ideas into production-ready technical solutions, balancing speed of execution with long-term maintainability - Built production AI pipelines combining RAG, tool calling, structured outputs (Pydantic), vector search (PGVector), prompt versioning, and Langfuse tracing - Engineered the real-time chat architecture using WebSockets with event orchestration, rate limiting, and low-latency communication - Built a multi-provider payment platform supporting one-time purchases, subscriptions, premium features, and monetization 🛠 Tech Stack Backend: Python, FastAPI, SQLAlchemy, Alembic, Pydantic Data: PostgreSQL, PGVector, Redis, RabbitMQ Infrastructure: Heroku, Docker, Cloudflare, RunPod, AWS Observability: Langfuse, New Relic, Sentry, PagerDuty AI: OpenAI, Google Gemini, Groq, ElevenLabs, self-hosted LLMs & Stable Diffusion
Played a key role in driving AI and automation initiatives at Sossego, designing and delivering production-grade solutions in LLMs, RAG systems, and hyper-automation for the insurance sector. 📊 Key Business Impact - Led the development of an AI-powered customer support agent (RAG + LLM) from scratch, delivering human-like assistance for insurance users in Brazil improving scalability and operational efficiency - Built a Data-as-a-Service (DaaS) platform from the ground up: - Owned critical parts of ETL pipelines (extraction, transformation, loading) - Developed data pipelines using Python and Jupyter Notebooks - Orchestrated workflows with Dagster - Structured ClickHouse as the central data warehouse 📊 Core Contributions - Designed and implemented a RAG-based AI agent using Python, ClickHouse, ClickHouse Vector Search, including prompt engineering and structured outputs - Built a high-performance search engine for a ticketing system using Node.js and PostgreSQL full-text search - Streamlined a document classification pipeline (contracts, policies, proposals) using Python and regex, orchestrated via Apache Airflow - Implemented an audio transcription pipeline using Whisper (a speech recognition model), enabling automated processing of voice data, also orchestrated with Airflow - Integrated ClickHouse as an analytical database and NATS as a messaging system, enhancing data processing and real-time capabilities 🛠️ Tech Stack - Languages & Frameworks: Python, FastAPI, Node.js, TypeScript, NestJS - Data & Storage: PostgreSQL, Redis, ClickHouse, MongoDB, BigQuery - Data Engineering & Orchestration: Airflow, Dagster, Temporal.io, Camunda - Messaging & Streaming: NATS, Google Pub/Sub - AI/ML: OpenAI, Embeddings, Vector Databases, Whisper - DevOps & Observability: ArgoCD, Prometheus, Grafana - Analytics & BI: Metabase
I worked on the Invoice-to-Pay platform, an end-to-end hyper-automation solution for tax document processing, transforming a highly manual workflow into a scalable and AI-driven (Google Document AI with trained models) pipeline. Streamlined a fully automated flow covering document ingestion, OCR-based classification, structured data extraction (OCR + RegEx), intelligent matching with service sheets and purchase orders, automated tax analysis pipeline with human-in-the-loop learning, and ERP integrations reducing the need for human intervention to rare edge cases. Delivered high-impact improvements across performance, cost, and scalability while ensuring reliability in a mission-critical tax environment. 📊 Key Achievements - Reduced manual effort in tax review processes by up to 90% by implementing a rule-learning system with a human-in-the-loop, where analyst-reviewed cases were persisted and automatically reused to process future documents with similar tax contexts. - Built and deployed a tax analysis module for CTE (Electronic Transport Invoices), expanding product coverage and compliance capabilities. - Cut Google Cloud costs by ~30% through optimization of infrastructure and processing pipelines. - Improved operational efficiency by reducing backlog and minimizing pending task volume. - Achieved a low rework rate (14%), ensuring high-quality deliveries across features and bug fixes. - Developed a smart search system using MongoDB Atlas full-text search, significantly enhancing response time, accuracy, and user experience. - Refactored the authenticity query system into an event-driven architecture, enabling high scalability and reliability across 200+ web crawlers. - Automated back-office workflows using Node.js and Golang scripts, reducing manual overhead and increasing operational speed.
As a Backend Developer I've worked on large-scale e-commerce projects including Reserva and Bosch hardwares. In addition to these client-facing projects, I've automated several internal processes for back-office operations. One notable automation I've created was designed to process large data ingestion anticipating client GO LIVES. Additionally I've developed a command-line tool for Front-end Developers by streamlining a setup for VTEX IO components which has provided a properly creation of email templates used in the e-commerce platform such as those triggered when a customer makes a purchase. This tool is still available on npm: https://www.npmjs.com/package/original-iostore. Through these efforts I also was able to optimize AWS costs by reducing the monthly bill from around $1,000 to just $180–$300 by month.
I’ve played a key role by migrating a legacy web system to a modern, client/server architecture. The migration has moved the system from outdated technologies (PHP 5, Apache, MySQL, HTML 4) to a way versatile, scalable and maintainable setup using Vue 2, Quasar Framework, PHP 7.4, Nginx, and Flight PHP for REST APIs which has significantly improved system performance and flexibility.
I’ve streamlined contract management for the New Business department by replacing Excel-based processes by a centralized contract system. The system has improved the identification of contracts nearing expiration saving time by improving anticipation actions for client renewals.
I’ve contributed to improve the company’s processes by developing automation tools for online media monitoring, enabling the daily processing of around 100,000 news articles through data crawling and scraping. I’ve also implemented text recognition automation to track text from images (OCR) and PDFs.