CV

Work Experience

NVIDIA AI Developer Advocate — Physical AI, Robotics & Open Models, Full-time, Remote, Barcelona, Spain, Aug 2025 – Present
- Build and teach advanced workflows for robotics and Physical AI, including simulation, teleoperation, synthetic data generation, functional behavior training, and human–robot–object interaction pipelines using Isaac Sim, Genesis, GR00T, NemoClaw, OpenShell, and ROS/ROS 2.
- Design, refine, and evaluate robot behaviors in simulation, deploying optimized and reliable policies on real hardware for high-performance real-world tasks.
- Develop and optimize real-time perception, manipulation, and control pipelines using efficient deep learning, ROS/ROS 2, TensorRT, CUDA, and hardware-aware inference for edge robotics platforms (Jetson Orin, Jetson Thor).
- Deploy, accelerate, and benchmark LLMs, VLMs, VLAs, and multimodal AI systems using TensorRT-LLM, vLLM, SGLang, and FlashInfer — enabling ultra-low-latency reasoning and task planning on Jetson and DGX platforms.
- Build scalable edge-to-cloud robotics AI architectures, integrating Jetson edge inference with cloud-native microservices for distributed compute, fleet management, monitoring, and CI/CD.
- Lead open-source robotics hackathons end-to-end, contributing hands-on to code, models, algorithms, and debugging while mentoring teams.
- Deliver high-impact workshops and live technical demos across the NVIDIA AI ecosystem — from Jetson to DGX Spark, DGX, and Blackwell (GB300) — driving developer adoption and ecosystem growth.
- Conduct applied research and advocacy on open models and inference, collaborating with framework teams (PyTorch, TensorFlow, ONNX Runtime, JAX, Triton) to optimize performance across diverse platforms.
- Champion best practices in quantization, graph optimization, kernel tuning, and hardware-aware model compression, influencing both internal roadmaps and external partners.
- Research and prototype social cognition and social behavior in robotics, enabling robots to understand human intent, cooperate naturally, and operate safely in shared environments.
Human Pose Recovery and Behavior Analysis (HuPBA), Universitat de Barcelona PhD Researcher, Hybrid, Barcelona, Sep 2020 – Present
- PhD research in Human-Robot-Object Interactions (HROIs) with Generative AI, at the intersection of computer vision, robotics, psychology, and neuroscience.
- Co-organized international challenges and seminars on gesture recognition and motion analysis (ChaLearn LAP).
- Specialized in multi-modal data analysis, human pose estimation, and non-verbal behavior forecasting.
- Annotated multi-modal features for the UDIVA dataset, enabling research in social behavior.
Fundació Bosch i Gimpera Deep Learning Developer, Hybrid, Barcelona, Nov 2023 – Sep 2024
- Developed YOLO-based multi-object detection and tracking for complex environments, achieving 85% detection precision.
- Applied knowledge distillation and quantization techniques for Edge AI, enhancing performance by 470%.
- International private project focused on small multi-object detection in challenging conditions.
Milestone Systems Computer Vision Research Engineer, Remote, Barcelona / Denmark, Mar 2023 – Nov 2023
- Authored a WACV 2024 paper on trajectory-based loitering detection for Smart Cities.
- Developed and validated geometric motion descriptors for robust loitering detection using thermal surveillance data.
Brodynt / Expereo Junior IT Manager, Hybrid, Barcelona, Aug 2020 – Sep 2021
- Managed over 1M devices, automating workflows with AI-driven anomaly detection.
- Designed Python-based monitoring systems to optimize device management.

Education

Universitat de Barcelona / Computer Vision Center Ph.D. in Computer Vision, Barcelona, May 2024 – Present (expected 2027) Focused on Human-Robot-Object Interactions (HROIs), Generative AI, and robotics, integrating psychology and neuroscience with AI technologies.
- Ported over 30 libraries for NVIDIA Jetson devices
- Research at the intersection of social cognition, multimodal models, and robotics
Universitat Autònoma de Barcelona / Computer Vision Center M.S. in Computer Vision, Barcelona, Sep 2023 Relevant Courses: Computer Vision, Video Analysis, 3D Vision, Visual Detection
Universitat de Barcelona M.S. in Fundamental Principles of Data Science, Barcelona, July 2022 Honors: Computer Vision
Universitat de Barcelona B.S. in Computer Engineering, Barcelona, June 2021 Honors: Machine Learning, Third Prize in Bachelor’s Thesis

Selected Open-Source Contributions

Enabling next-gen NVIDIA hardware across the open-source AI stack:

Repository	PR	Description
Dao-AILab/flash-attention	#1904	CUDA 13 + sm12x support
Dao-AILab/flash-attention	#2222	Hybrid Flash Attention
vllm-project/flash-attention	#95	CUDA 13 for vLLM FA
sgl-project/sglang	#11299	Enable Thor/Spark/GB300
sgl-project/sgl-flash-attn	#8	SGLang Flash Attention CUDA 13
pytorch/pytorch	#165048	Fix CUDSS for Thor/Spark
facebookresearch/xformers	#1344	Blackwell support
state-spaces/mamba	#776	Mamba CUDA 13
bitsandbytes-foundation/bitsandbytes	#1491	Blackwell binaries
opencv/opencv	#27537	Refactor Blackwell
kvcache-ai/Mooncake	#344	Enable SBSA

Robotics & Edge AI:

Repository	PR	Description
NVIDIA-AI-IOT/jetson-containers	#258	Isaac Sim, Isaac Lab, Newton, Warp, CuPy, Numba
NVIDIA-AI-IOT/jetson-containers	#240	Jetson Thor GR00T + LeRobot
dusty-nv/jetson-containers	#1391	SGLang, vLLM, TF, Triton Blackwell
NVIDIA/Isaac-GR00T	#212	decord2 integration
NVIDIA-AI-IOT/jetson-ai-lab	#343	Cosmos Reasoning 2B
rbonghi/jetson_stats	#718	Fix jtop for Thor/Spark

Own projects: decord2 (47+ stars), edge2cloud, DGX Spark Playbook

Industry impact: Contributions to enabling the Grace/Hopper (GH200) architecture helped major cloud GPU providers fully sell out their GH200 fleet capacity, demonstrating the direct business value of making new hardware platforms accessible to the developer community.

Skills

Programming Languages: C/C++, Python, CUDA, ROCm, R, Java, SQL, Bash, YAML
AI Frameworks: PyTorch, TensorFlow, JAX, ONNX Runtime, TensorRT, TensorRT-LLM, vLLM, SGLang
Robotics: ROS/ROS 2, Isaac Sim, Isaac Lab, GR00T, MoveIt2, Genesis
Edge & HPC: NVIDIA Jetson (Orin, Thor), DGX Spark, DGX, Blackwell (GB300), CUDA optimization
Infrastructure: Docker, Kubernetes, Slurm, CI/CD, cloud-native microservices
Tools: Git, LaTeX, NVIDIA Nsight, Hydra, Google Cloud, AWS

Languages

Catalan — Native
Spanish — Native
English — Full professional proficiency
Chinese (Mandarin) — Currently learning

Publications

The Fused Kernel Library: A C++ API to Develop Highly-Efficient GPU Libraries

O. Amoros, A. Andaluz, J. Nunez, A.J. Pena. (2025). "The Fused Kernel Library: A C++ API to Develop Highly-Efficient GPU Libraries." arXiv preprint arXiv:2508.07071.

Enhancing Clinical Psychology Practice through Data-Driven Machine Learning Monitoring Systems

G. Martínez, A. Trujillo, J. Núñez, J.C.S. Jacques, A. Clapés, et al. (2025). "Enhancing clinical psychology practice through data-driven machine learning monitoring systems."

Identifying Loitering Behavior with Trajectory Analysis

Johnny Núñez, Zenjie Li, Sergio Escalera, Kamal Nasrollahi. (2024). "Identifying Loitering Behavior with Trajectory Analysis." Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). 251-259.

Comparison of Spatio-Temporal Hand Pose Denoising Models

Johnny Núñez. (2022). "Comparison of Spatio-Temporal Hand Pose Denoising Models." Universitat de Barcelona.

Chalearn LAP challenges on self-reported personality recognition and non-verbal behavior forecasting during social dyadic interactions: Dataset, design, and results

Cristina Palmero, German Barquero, Julio CS Jacques Junior, Albert Clapés, Johnny Núnez, David Curto, Sorina Smeureanu, Javier Selva, Zejian Zhang, David Saeteros, David Gallardo-Pujol, Georgina Guilera, David Leiva, Feng Han, Xiaoxue Feng, Jennifer He, Wei-Wei Tu, Thomas B Moeslund, Isabelle Guyon, Sergio Escalera. (2022). "Chalearn LAP challenges on self-reported personality recognition and non-verbal behavior forecasting during social dyadic interactions: Dataset, design, and results." Understanding Social Behavior in Dyadic and Small Group Interactions. 4-52.

Comparison of Spatio-Temporal Models for Human Motion and Pose Forecasting in Face-to-Face Interaction Scenarios: Supplementary Material

German Barquero, Johnny Núnez, Zhen Xu, Sergio Escalera, Wei-Wei Tu, Isabelle Guyon, Cristina Palmero. (2022). "Comparison of Spatio-Temporal Models for Human Motion and Pose Forecasting in Face-to-Face Interaction Scenarios: Supplementary Material."

Didn’t see that coming: a survey on non-verbal social human behavior forecasting

German Barquero, Johnny Núnez, Sergio Escalera, Zhen Xu, Wei-Wei Tu, Isabelle Guyon, Cristina Palmero. (2022). "Didn’t see that coming: a survey on non-verbal social human behavior forecasting." Understanding Social Behavior in Dyadic and Small Group Interactions. 139-178.