Nakato
Research Scientist
Focused on Multimodal Representation Learning and Large Language Models. Bridging the gap between vision and language through scalable architectural innovations and RLHF.
About Me
I am a dedicated researcher with a passion for understanding how machines can learn more cohesive representations of our world. My work currently explores the intersection of visual perception and linguistic reasoning.
- check_circle Multimodal Alignment & Contrastive Learning
- check_circle Efficient Attention Mechanisms in Transformers
- check_circle Human-in-the-loop Reinforcement Learning (RLHF)
- check_circle Representation Alignment across Domains
Featured Research Projects
Imitator
A multimodal framework for gesture-to-speech translation using transformer architectures. Focused on zero-shot generalization across diverse speaker profiles.
ActivAdda
Implemented activation addition (steering vectors) to bias LLM responses without parameter updates.
SKAI - Affective Response Generation
Trained a reward model over fine-grained emotional categories and adapted GPT-2 to generate supportive and sentiment-aware responses.
Experiencia Laboral
Superintendencia de Banca, Seguros y AFP
presencialAI Developer
Automation of end-to-end testing with Cypress, managing data pipelines and Jira integration for large-scale deployments.
Educación
Pontificia Universidad Católica del Perú
Artificial Intelligence (AI) Summer Camp
Universidad Peruana de Ciencias Aplicadas
B.S. Computer Science
Specialization in Artificial Intelligence and Neuroscience
Selected Publications
Learning Motion-Based Embeddings for Sign Language Retrieval.
Manuscript in preparation (Draft) — 2026
Imitator: Multimodal Sign Language Model
SimBig25 International Conference on Information Management and Big Data — 2025
Let's build the future.
Interested in collaborating on Data/ML research? I'm always open to discussing new opportunities or academic partnerships.