Available for CV / ML work
Luís Condados
Computer vision and ML engineer. I build open-source tools for perception — detection, segmentation, OCR, VLMs — with a bias toward small, sharp APIs over framework sprawl.
Now
Building acaua —
an Apache-2.0, PyTorch-native computer vision library. One API
(predict / train / export) across detection, segmentation,
OBB, OCR, VLMs, and zero-shot.
The goal: a pythonic alternative to mmdet and PaddleDetection — fewer abstractions,
less YAML, friendlier extras like acaua[ocr], acaua[vlm],
acaua[zeroshot], acaua[export].
Featured
PyTorch-native CV library with a single unified API. One class to learn, one to remember.
Locating lung opacities on chest radiographs — detection on a real medical imaging benchmark.
Lateral and longitudinal controllers, full self-driving stack built during the Coursera Self-Driving Cars specialization.
Reproduces a target image using only black thread on white — exhaustive search over line configurations to approximate tone and form.
Full software stack for UFRN's Poti robot soccer team — vision, AI, control, Bluetooth comms, custom V4L2 camera layer.
Raytracing renderer in C++ with OpenGL for rasterization and OpenMP for parallel pixel shading.
Selected work
Background
Engineering background from UFRN (DCA) with a thesis on embedded control for differential-drive robots. Years of hands-on robotics work with the Poti team — computer vision, control, embedded C. Now focused on making modern CV models actually usable in production Python.