Summary
Victor Shirasuna is an AI research scientist and MSc student in Computer Science at IME-USP with nine years of experience building and scaling large multimodal and language models across HPC environments. He has hands-on expertise training models like LLMs, VQGAN and CLIP with PyTorch DDP and DeepSpeed on hundreds of GPUs in cloud and national lab clusters, and has automated end-to-end validation pipelines for materials-focused foundation models at IBM Research. Victor’s work spans both applied engineering and peer-reviewed research, with publications at top AI workshops (NeurIPS, ICML, ICLR, AAAI) and projects targeting molecular and materials discovery. Based in São Paulo, he brings a rare combination of production-grade distributed training experience and domain knowledge in chemistry-focused ML, and maintains an active GitHub (vichShir) evidencing his ongoing tooling and reproducibility efforts.
9 years of coding experience
Advanced (Level 13) English Program, Advanced (Level 13) English Program at ILAC
Bachelor's degree Information Systems, Bachelor's degree Information Systems at Universidade Estadual de Campinas
Master of Science - MS Computer Science, Master of Science - MS Computer Science at Instituto de Matemática, Estatística e Ciência da Computação (IME-USP)
Escolas Técnicas Estaduais
English, Portuguese