Looking for a job

Giovanni Trabucco

@giovannitrabucco

I build embedded AI and real-time software for LiDAR perception and radar simulation.

At Tattile, I develop convolutional neural networks for classifying LiDAR-generated 2D vehicle point clouds, deploying them on CUDA-less single-core ARMv7 machines with average inference times below 100ms.

Previously, I built real-time software for Leonardo's RADAR simulations supporting OCCAR's European MALE RPAS project and researched gradient-free deep learning methods for audio pre-compensation at Sony. I'm a manual memory management enthusiast currently fixing the bugs I introduced in the last commit.

Experience

Work history, roles, and key accomplishments

Current

Software Engineer

Current

Tattile srl

Jul 2024 - Present (2 years 2 months)

Development of Convolutional Neural Network architectures for classification of LiDAR-generated 2D vehicle point clouds and deployment on CUDA-less, single core (ARMv7 32-bit) machines under strict time constraints (average per-sample inference time < 100ms).

Software Engineer

Leonardo

Mar 2023 - Jun 2024 (1 year 3 months)

Development of real-time software in the context of RADAR simulations targeting OCCAR’s “European MALE RPAS” project.

Research Intern

Sony

Jul 2022 - Jan 2023 (6 months)

Deep learning for gradient-free development of pre-compensation systems for black box audio models' frequency response via Simultaneous Perturbation Stochastic Approximation (SPSA)-like methods.