Same cool projects, possibly cooler

Same cool projects.

Leading the work on the framework that runs inference of the deep learning networks that power the face and scene recognition features in Photos. Also powering the newly announced CoreML.All inference happens locally on device with minimal footprint and exploiting all the efficiency provided by the A silicon.

The main topics of my research activity are:- Development of efficient numerical tools for simulating laser-plasma accelerators. I've implemented a fully-3D, electromagnetic Particle-in-Cell code which runs and scales well on GPU clusters. This work awarded me the Student presentation prize, Computational working group, at “15th Advanced Accelrators Concepts Workshop”, Austin, TX 2012 (see "Honors and Awards" section and http://proceedings.aip.org/resource/2/apcpcs/1507/1/184_1).- Providing simulations for the INFN-SPARCLAB experiments G-Resist, SITE and LILIA.

Personal project. An iPhone camera app that provides funny effects and seamless photomontages, all implemented using computer vision algorithms.http://wonderc.am

I've been working on a Quasi-Static, 2D cylindrical, Particle in cell code for simulating Laser Wakefield Electron Acceleration: INF&RNO quasi-static.I've improved the convergence of the iterative numerical scheme and implemented parallelization using pipelining techniques and dynamic load balancing for achieving scalability in situations in which the convergence rate of numerical scheme varies spatially.

I've been working on improving the performance of the NVIDIA's GPU multi grid library NVAMG (the core of ANSYS Fluent GPU module).

I've implemented an intelligent digital steering player assistance and started the implementation of racing bots.

I've been working on, 2D cylindrical, Particle in cell code for simulating Laser Wakefield Electron Acceleration: INF&RNO.- I've written a visual tool for visualizing the code's results using cylindrically symmetric volume rendering: https://dl.dropboxusercontent.com/u/7256385/ad/thesis_particle_in_cell/bubble.avihttps://dl.dropboxusercontent.com/u/7256385/loasis2.avi - I've parallelized the code using a non-trivial, iterative technique for solving the implicit envelope equation in parallel.

I've implemented a real time renderer for the virtual heritage project "NUME". The renderer precomputes (via GPU CUDA raytracing) indirect illumination data and stores it in spherical harmonics coefficient textures.https://dl.dropboxusercontent.com/u/7256385/ad/procedural_world/Nume2009_1min_pomeriggio.mpvhttp://code.google.com/p/nume/