Thiago Trannin

- Assessed the impact in customer conversions of using event data from the company's in-house e-storefront platform. Implemented integrations with the Conversion APIs from Meta and TikTok. - This position was sponsored by the Japanese government by means of a coding contest which involved students from countries in Africa and Latin America. The best candidates were invited for a month-long internship at a Japanese company.

- Developed embedded software that runs in traffic control cameras. - Worked on very diverse tasks in firmware development and systems programming: from optimization of image processing routines and design of data processing pipelines to parsers, logging systems, FFI, network protocols, CPU and memory profiling, web interfaces, and automated testing. - Refactored a monolithic C++ project in smaller, independent modules. Pioneered Rust adoption in safety-critical functionality and maintained its integration with existing code via the C ABI.

- Designed a keypoint detection model for estimation of the position in 3D space of an excavator machine's arm; the model's input was the video feed from a camera above the operator's seat. - This position was sponsored by the Japanese government by means of a coding contest which involved students from countries in Africa and Latin America. The best candidates were invited for a month-long internship at a Japanese company.

- Applied classic computer vision algorithms to detect vehicles stopped over crosswalks. - Researched and developed deep learning models for smart traffic monitoring applications, such as vehicle classification, vehicle tracking and license plate reading via optical character recognition. - Deployed cross-compiled native libraries that abstracted over the aforementioned problems, with accompanying wrappers for high-level languages.

- Applied classic computer vision algorithms to detect vehicles stopped over crosswalks. - Researched and developed deep learning models for smart traffic monitoring applications, such as vehicle classification, vehicle tracking and license plate reading via optical character recognition. - Deployed cross-compiled native libraries that abstracted over the aforementioned problems, with accompanying wrappers for high-level languages.