M.S. Computer Science at the University of Trento (expected March 2027). GPU, HPC and distributed systems. EU citizen, available March 2027.
dist-patoh: a distributed multi-GPU hypergraph partitioner in C++20, CUDA and MPI. It splits massive, irregular sparse graphs across GPUs while balancing compute and minimising cross-node communication. The multilevel pipeline covers parallel coarsening, recursive k-way partitioning and local refinement. Code private during thesis writing; public release after submission.
Software Engineer Intern at Bitmovin (Jun–Sep 2026). Co-built KAIROS, AI highlight detection and segmentation for sports, news and podcasts: live/VOD, sport understanding, orchestration, and a public MCP API.
- cuda-SpMV: hybrid adaptive A30 kernel combining thread-per-row and warp-per-row CSR; diagnosed the overscheduling failure that motivated the thesis
- parallel-mst: MPI+OpenMP Borůvka MST with a flat adjacency matrix and mmap I/O; measured 2-32 rank scaling on one 64-core node
- distributed-storage-system: Akka/Java 21 key-value store with consistent hashing, N/W/R quorums and node-failure recovery
- DWT-SVD-watermarking: 1st of 11 teams, CTM Challenge 2025
- mpeg-dash-analysis: bachelor's thesis: MPEG-DASH under simulated (Mininet) and real (AWS) networks
C, C++17/20, CUDA, Go, Python, Java | MPI, OpenMP, Slurm, CMake, cuSPARSE | Linux, Docker, Terraform, GCP, PostgreSQL, Git



