Pierluigi Urru

Pierluigi joined Microsoft’s Applied Sciences Group (ASG) as a senior applied scientist, working on A.I.

Before joining ASG, Pierluigi worked at Arm, on the Machine Learning Model Conditioning team where he contributed to the TensorFlow Model Optimization Toolkit (TF-MOT), used to shrink TensorFlow models and speed up inference while preserving accuracy, and the Tensor Operator Set Architecture (TOSA), an open standard for tensor operations maintained by Arm which is independent of any hardware or software design.  Pierluigi also worked in the Fast Models team, where he developed performance models for upcoming GPU, NPU, ISP, and automotive IP offerings.

Before Arm, Pierluigi was at Visa, where he worked on a project to re-architect the company’s payment processing system as a distributed system using open technologies.  Before Visa, he interned at Thomson Reuters, where he conducted big-data analytics on foreign-exchange trading data.

Pierluigi’s fascination with machine learning began during his university studies.  His thesis involved implementing various machine-learning models for music composition, and evaluating each model’s strength on criteria such as originality, song structure, harmony, rhythm, and more.  After graduation he continued to expand his machine-learning knowledge through online courses.  His current areas of interest include LLMs and robot learning.

Outside of work Pierluigi is most passionate about playing and making music.