Lucas Pires Camargo

At Luxoft, I worked on the ADAS software stack for a next-generation software-defined vehicle (SDV) developed by a major European OEM.
My role was in the Adaptive AUTOSAR service layer, which was responsible for bridging the vehicle's internal ADAS logical core, implemented in MATLAB/Simulink via model-based engineering, and the broader vehicle software ecosystem. This included communication with a 3rd-party automotive vision and feature stack, as well as managing activation and deactivation logic, overrides, and precondition handling across vehicle subsystems developed by multiple independent teams.

The core of my work was the C++17 wrapper layer that translated the Simulink-generated logical core into compliant Adaptive AUTOSAR interfaces, ensuring correct integration with the surrounding vehicle processes. Where the model-based approach proved infeasible or overly complex, I implemented the logic directly in handwritten C++.

The project operated under Automotive SPICE, which imposed rigorous process requirements, including test-driven development and full requirements traceability throughout the development cycle.

At Luxoft, I worked on the UI layer of a next-generation automotive instrument cluster platform for a major European OEM, developed entirely in Qt and QML with a custom component library.

My responsibilities centered on UI bring-up and interface implementation, bringing up new screens and visual components from specification to working software on the target hardware. This included work on the heads-up display interface, as well as ongoing bug fixing and visual polish across the broader cluster UI.

The platform operated under AUTOSAR Adaptive, which meant UI development had to account for strict real-time and safety constraints alongside the usual challenges of performance and visual fidelity on automotive-grade hardware. Development was carried out entirely on Linux, with the team working against an automotive-grade target runtime environment.

Pollux Automation uses computer vision systems for different use cases in industrial settings. Some of these require a software-defined approach built on a digital manufacturing system (DMS) platform.

My role was to lead the team in charge of developing this CV module of the DMS, assist with hardware integration of IP cameras, establish the platform facilities and abstraction layers for assembling image processing pipelines, retrofitting and improving existing ML workflows into the vision platform, and defining the UI/UX paradigms for configuration, training, dataset management, and production flows.

Building on top of a Python codebase, we integrated proven ML frameworks such as PyTorch. We developed custom CFFI (C/C++-based) integrations for relevant software components, optimizing for predictable real-time performance, efficiency, and reliability.

At Bluechip Studio, a small casino games studio based in Joinville, Brazil, I was in charge of modernizing a codebase that was over 10 years old while also maintaining and customizing the Linux distribution it ran on.

I served as the company's Linux and C++ platform specialist, ensuring that content programmers had the tools they needed and that the platform ran well on up-to-date hardware. The role was broad by nature, and I ended up building a significant amount of custom tooling along the way.

The most technically interesting part of the work was rewriting the entire rendering stack from scratch in OpenGL ES, replacing a legacy OpenGL 1.x codebase that had accumulated over a decade of additions. This brought substantial performance improvements and, crucially, made the platform portable to embedded hardware such as ARM SBCs.

Maintaining API and behavioral compatibility with existing games was the primary constraint, but the new foundation also opened the door to capabilities previously unavailable, including particle systems and 3D rendering layers.

This is a personal project. I built a multiplatform game engine with all the functionality required to implement 2D games. In addition to hardware-accelerated rendering and a node-based audio processing system, there are facilities for physics, tilemaps, particle systems, and more. A custom Lua-based scripting system ties it all together. Linux, Windows, Android, the Web, and even Nintendo Wii builds have been tested.

I work on this occasionally when time allows. It's first and foremost a fun thing for me to do.