Jiaming Song

Led research and co-authored 'eDiff-I: Text-to-Image Diffusion Models with Ensemble of Expert Denoisers' (arXiv:2211.01324), contributing to state-of-the-art text-to-image generation and influencing NVIDIA's generative AI roadmap.

Developed and optimized novel diffusion implicit models, enhancing the efficiency and quality of generative processes for real-world applications.

Collaborated cross-functionally to translate complex research breakthroughs into practical solutions, impacting future product development cycles.

Contributed to the strategic direction of AI research, identifying key areas for innovation in large-scale model development and deployment.

Pioneered research on 'A General Recipe for Likelihood-free Bayesian Optimization,' which received a long oral presentation (Top 2.2%) at ICML 2022, demonstrating significant advancements in optimization techniques.

Published multiple influential papers in top-tier conferences, including NeurIPS 2022 and AAAI 2023, advancing the theoretical and practical aspects of diffusion models and imitation learning.

Mentored and guided junior researchers and graduate students, contributing to their academic growth and successful project completion.

Authored 'Comparing Distributions by Measuring Differences that Affect Decision Making,' which earned the ICLR 2022 Outstanding Paper Award, a top recognition in the field.

Developed Denoising Diffusion Implicit Models (DDIMs), a seminal contribution to diffusion models that significantly improved sampling speed and quality.

Secured the prestigious Qualcomm Innovation Fellowship for developing 'Safe Multi-Agent Imitation Learning for Self-Driving,' showcasing innovative application of AI.

Published over 20 peer-reviewed papers in leading AI/ML conferences (NeurIPS, ICML, ICLR, AAAI, ECCV), establishing expertise in core AI methodologies.

Developed and implemented novel algorithms for 'Large-scale Object Counting from Satellite Images,' enhancing the accuracy and efficiency of geospatial analysis.

Collaborated effectively within a world-class research team to push the boundaries of generative models and computer vision.

Presented research findings to senior scientists, contributing to the team's intellectual property and strategic initiatives.

Contributed to the development of 'Learning Interpretable Skill Abstractions from Language (LISA),' enhancing the interpretability and understanding of complex AI behaviors.

Assisted in designing and executing large-scale experiments for advanced AI models, optimizing performance and scalability.

Collaborated on research initiatives that significantly advanced the state-of-the-art in reinforcement learning and generative modeling.

Contributed to the development of robust image classification models, improving recognition accuracy for various real-world scenarios.

Implemented and evaluated diverse deep learning architectures, gaining practical experience in model selection and optimization.

Applied theoretical AI concepts to real-world challenges in facial recognition and image analysis, enhancing product capabilities.

Explored and applied statistical modeling and machine learning techniques to analyze complex datasets, laying groundwork for future research.

Collaborated on a research project focused on Factored Temporal Sigmoid Belief Networks for sequence learning, contributing to early insights.

Developed a strong understanding of research methodologies, experimental design, and scientific writing.