Matheus – React, Typescript, .NET

Internal tooling for back office automation. The agentic system interprets documents and automates internal processes with a Human-in-the-loop setup via Microsoft Teams.

The agent is responsible for:

• Booking invoices into the accounting tool
• Generating payment files
• Booking receipts into the accounting tool
• Booking reimbursements into the accounting tool

For privacy reasons, the LLMs used in this project are self-hosted.

• System architecture for the automation
• Implement system prompts for the agents
• Implement human-in-the-loop flow via Teams with feedback and approval
• Test the system with multiple documents
• Evaluate different open models
• Deploy the automation and model on the Cloud (Azure)

It's self-checkout systems for multiple convenience brands across Europe.

• Implemented the MVP React application and central state management.
• Built the promotion engine library in .NET.
• Architected a central payment system for all stores.
• Wrote integration and end-to-end tests.
• Optimized database queries for better performance.
• Automated infrastructure provisioning and deployment with Terraform.
• Mentored and onboarded 5+ developers to the project.
• Coordinated with internal and external teams, including security and auditors, to ensure compliance for self-checkout systems.

It's a secure messaging layer that enables Swiss organizations to exchange sensitive data reliably and in a structured way. It provides API and webhook-based integration for seamless message delivery, processing, and auditing across systems.

• Designed and implemented a secure messaging architecture for the Swiss company network.
• Built API- and webhook-based communication flows for reliable message exchange.
• Developed Blazor SSR pages to administer message delivery.
• Implemented event sourcing for full auditability and traceability.
• Built resilient message processing with retry, queuing, and failure handling.
• Designed and optimized PostgreSQL data models for sensitive, high-integrity data.
• Applied end-to-end security practices for confidential information handling.
• Automated deployment pipelines to a private Kubernetes cluster.
• Designed CI/CD workflows for versioning, rollout, and rollback.
• Containerized services and managed Kubernetes orchestration.
• Implemented logging, monitoring, and observability for distributed message flows.

It's an open-source multi-currency budgeting and finance tracker, designed to help users manage budgets, expenses, and assets.

• Implemented multi-currency financial models and calculations.
• Built a scalable full-stack architecture with Next.js, Drizzle ORM, and PostgreSQL.
• Developed dynamic reporting and expense visualization features.
• Maintained code quality, testing, and open-source documentation.
• Implemented MCP server for agentic data fetching

It's a secure document exchange platform that enables structured, compliant communication between institutions, allowing them to efficiently request information and share updates related to insured individuals and companies.

• Designed and implemented a full system architecture covering front-end, back-end, and infrastructure.
• Modeled and optimized database schema for structured document exchange.
• Built scalable APIs and core business logic for inter-agency workflows.
• Developed modular, reusable frontend component system.
• Implemented advanced search and filtering across documents and entities.
• Automated infrastructure provisioning and deployment with Terraform.
• Ensured compliance with D-SIPD standards and secure data exchange protocols.
• Implemented end-to-end testing strategy (unit, integration, and e2e).
• Optimized performance, reliability, and system observability.
• Led debugging, issue resolution, and ongoing system improvements.

It's a digital receipt platform that allows customers to receive receipts instantly via QR codes.

• Designed and implemented an end-to-end workflow for digital receipt delivery.
• Built a secure QR-code system linking transactions to individual customers.
• Implemented real-time receipt delivery via server-sent events.
• Developed a fast, resilient back-end for transaction feeds and user requests.
• Created front-end pages for instant receipt viewing and safe reloads.
• Ensured transaction integrity and accurate QR-to-receipt mapping.
• Integrated with the company's point-of-sale systems for seamless rollout.
• Added logging, monitoring, and error handling for reliable delivery.

Central Analytics Ingestion is a serverless data pipeline platform designed to reliably ingest, process, and distribute events across the company. Built with Node.js and AWS services (Lambda, SQS, SNS, DynamoDB, CloudFormation), the system handles high-volume event streams with deduplication, real-time processing, and scalable delivery to multiple internal consumers.

• Architected the end-to-end serverless ingestion and publishing system for company-wide event distribution
• mplemented Node.js Lambda functions to process DynamoDB streams, deduplicate events, and publish to internal consumers
• Designed and deployed infrastructure using CloudFormation with automated CI/CD pipelines via Jenkins
• Onboarded and mentored a junior engineer, transferring knowledge of serverless patterns and AWS best practices
• Optimized event processing for reliability, scalability, and low latency;
• Implemented monitoring, logging, and error handling to ensure observability and maintainability

It is a serverless analytics service built for performance views and financial charts. It consumes event streams from the Central Analytics Ingestion platform and computes financial metrics such as Internal Rate of Return (IRR), enabling real-time insights for reporting and visualization systems.

• Built a dedicated analytics consumer for the charts and performance views team.
• Implemented AWS Lambda functions to process Kinesis event streams in real time.
• Developed IRR calculation logic together with domain experts.
• Ensured idempotent processing and data consistency across reprocessed event streams
• Optimized Lambda execution for low-latency financial computations at scale and reduced cold-starts.
• Collaborated with domain experts and engineering teams to ensure correct calculations.