● deployed
LangWatch OSS
Open-source contributions to the LLM evaluation and AI agent testing platform. Improved curation interface and Redis Cluster integration for high availability.
↗ github.com/langwatch/langwatch
Status
Available for Work
Years of Experience
4+
Languages
🇧🇷 | 🇺🇸 | 🇯🇵
Location
Brazil
Professional Experience
Compass UOL — Software Engineer | Studio Data & AI
May 2022 → PresentDigital Bank — Top 5 BR
- Implementation and maintenance of conversational flows for the financial sector and POS terminals
- Architectural improvements in backend for high traffic across WhatsApp, website, and mobile app
- Reduced operational costs vs. human service through virtual agent implementation
- Monitoring via LangWatch platform for conversation curation
Virtual Agent Architecture — RAG + Observability
Intelligent virtual agent with flow orchestration via Langflow, Python microservices, vector knowledge base (PgVector) and full observability via LangWatch.
User Interface
WhatsApp
Web Chat
App
▼
Flow Orchestration
Langflow (Flow Engine)
Knowledge Base (RAG)
▼
Backends (Python)
Financial Agent
POS Agent
Support Agent
Internal APIs
▼
Observability & Monitoring (LangWatch)
Traces & Spans
Guardrails
Evaluations & Monitors
Cost & Latency Analytics
▼
Data & Storage
PostgreSQL
PgVector (Embeddings)
Redis (Cache)
Kubernetes
Python
Langflow
LangWatch
PostgreSQL
PgVector
RAG
OpenTelemetry
Redis
K8s
FastAPI
Chatbot Architecture — High-Scale Multichannel
Distributed conversational platform serving multiple channels (WhatsApp, Web, App) with asynchronous processing via event-driven architecture.
Channels
WhatsApp
Web
App (iOS/Android)
▼
Gateway & Routing
Kong API Gateway
WebSocket / Socket.IO
▼
Processing & NLU
Rasa (NLU + Dialogue)
Custom Actions (Python)
Genesys Cloud (Handoff)
▼
Event Streaming
Apache Kafka
Consumers (Node.js)
▼
Data & Infrastructure
MongoDB (Cluster)
Redis (AWS Cluster)
AWS S3
Kubernetes
Python
Node.js
TypeScript
Kafka
Kong
Rasa
Genesys
MongoDB
Redis
S3
K8s
Socket.IO
Airflow
Financial Company — RPA Tech Lead
- Technical leadership of the RPA (Robotic Process Automation) team
- Bot development with BotCity for desktop, web, SAP automation and API integrations
- Execution on Windows VMs with centralized orchestration
- Reduction of manual tasks and operational efficiency gains in banking back-office
RPA Architecture — Orchestrated Automation
Robotic process automation platform with BotCity as central orchestrator, running bots on Windows VMs for desktop, web, SAP and API automation in the banking back-office.
Target Platforms
Desktop Apps
Web Apps
SAP (GUI/Fiori)
REST APIs
▼
Orchestration
BotCity Maestro
Scheduler & Queue
Dashboard & Logs
▼
Execution (Python Bots)
Desktop Bot (PyAutoGUI)
Web Bot (Selenium)
SAP Bot (Scripting)
API Bot (Requests)
▼
Infrastructure & CI/CD
Windows VMs
Jenkins (CI/CD)
Artifact Registry
Python
BotCity
Windows VMs
Jenkins
SAP
Selenium
REST APIs
Desktop Automation
Customer Service Center — BPO Company
- Rasa framework for internal chatbots and automation of repetitive demands
- Implementation of the Rocket.Chat contact center platform
Contact Center Architecture — Omnichannel + NLU
Internal service platform with Rocket.Chat as omnichannel contact center, Rasa for natural language understanding (NLU/NLP), and Google Workspace for app and user management.
Service Channels
Internal Chat
Email
Live Chat (Widget)
▼
Omnichannel Platform (Rocket.Chat)
Rocket.Chat Server
Omnichannel Routing
Agent Queue & SLA
▼
NLU & Dialogue (Rasa)
Tokenizer + Featurizer
Intent Classifier (DIET)
Entity Extractor (CRF)
Dialogue Policies (TED)
▼
Management (Google Workspace)
Admin Console (SSO)
User Provisioning (OUs)
App Management
▼
Infrastructure
MongoDB
Python (Rasa Server)
Node.js (Rocket.Chat)
Rocket.Chat
Rasa
Python
Google Workspace
MongoDB
Node.js
NLU/NLP
Omnichannel
Projects
● deployed
Inithub — Collaborative Initiatives Platform
Full-stack monorepo for collaborative initiative management with AI support. React + Vite frontend, NestJS backend with Prisma, FastAPI AI agent with OpenAI, and PostgreSQL + pgvector for vector search.
↗ github.com/engsoft-pantanal-dev/inithub● deployed
Pantanal.dev — Fraud Detection
Visualization dashboard and API for fraud detection in financial transactions. ML models (Random Forest) with balancing techniques (SMOTE, ADASYN, Borderline-SMOTE) served via Flask/Dash and stored on AWS S3.
↗ github.com/Time-Jaguatirica/fraud-detection-apiEducation
Universidade Federal de Mato Grosso do Sul
Management Information Systems
AI, Decision Support Systems, Software Engineering, Database Lab, Software Quality, etc
Main Stack
Python
JavaScript
Artificial Intelligence
APIs / Microservices
SQL
Git
AWS
Kubernetes
React
Jenkins