Tony A. Ginart

Research Interests

I am broadly interested in artificial intelligence, cybernetics, and information science & engineering. My doctoral research is on theory and algorithms for large-scale machine learning. I work on making ML systems more efficient, scalable, secure and easier to deploy. 

Experience

• Co-Founder, Dialect (YCombinator) (2022 - 2023)

  • Building the future of generally reliable and capable LLM agents for complex process automation.

• Graduate Researcher, Stanford University (2017 - 2022)

  • Invented deletion-efficient data management algorithms for unsupervised learning (100× speedup)

  • Invented minimax rate optimal policy for ML deployment monitoring based on expert supervision (>25% increase in  label efficiency)

• Part-time Student Researcher, Facebook AI (Fall & Winter 2021)

  • State-of-the-art privacy-preserving protocols for large-scale transformer models (>1000× more privacy per user for given model quality)

• Research Intern, Facebook AI (Summer 2019, Summer 2021)

  • Designed embedding architecture for deep recommendation models that uses 16× fewer parameters and trains 3× faster without accuracy loss

  • Open-source contributor to facebookresearch/dlrm (over 3K stars on GitHub)

• Technical Staff Intern, Johns Hopkins University Applied Physics Laboratory (Summer 2017)

  • R&D for RL-based control algorithms with application to air & missile defense

• Undergraduate Researcher, UC Berkeley Laboratory of Information and Systems Science (Summer 2016) 

  • Invented state-of-the-art compression algorithm for genomic data (6× better than gzip)

• Undergraduate Researcher & Teaching Assistant, Washington University in St. Louis (2015-2016)

  • Developed a real-time NLP-based ML classifier for social media streams with application to public health monitoring

• Software Development Intern, Answers.com (Summer 2014)

  • Full-stack web development: LAMP stack backend & JS frontend

  • Fixed dozens of bugs on production site and added several features

Publications and Manuscripts

A.A. Ginart, L. van der Maaten, J. Zou, C. Guo. SubMix: Practical Private Prediction for Large-Scale Language Models. Preprint, 2022.

A.A. Ginart, M. Zhang, J. Zou. MLDemon: Deployment monitoring for machine learning systems.  International Conference on Artificial Intelligence and Statistics (AISTATS), 2022. Challenges in Deploying and Monitoring Machine Learning Systems @ ICML, 2021.

A.A. Ginart, M. Naumov, D. Mudigere, J. Yang, J. Zou. Mixed Dimension Embeddings with Application to Memory-Efficient Recommendation Systems. International Symposium on Information Theory (ISIT), 2021. PeRSonAl @ ISCA, 2020. Github.

A.A. Ginart, E. Zhang, Y. Kwon, J. Zou.  Competing AI: How does competition feedback affect machine learning?.  International Conference on Artificial Intelligence and Statistics (AISTATS), 2021. CoopAI @ NeurIPS, 2020. Press: Stanford HAI. 

S. Feizi,  F. Farnia, T. Ginart, D. Tse. Understanding GANs in the LQG Setting: Formulation, Generalization and Stability. IEEE Journal on Selected Areas in Information Theory, 2020.

A.A. Ginart, M. Y. Guan, G. Valiant, J. Zou. Making AI forget you: Data deletion in machine learning. Advances in Neural Information Processing Systems (NeurIPS), 2019. Spotlight. Github. Press: The Register, IEEE Spectrum.  

A.A. Ginart*, J. Hui*, K. Zhu*, I. Numanagic, T.A. Courtade, S.C. Sahinalp, D.N. Tse. Optimal compressed representation of high throughput sequence data via light assembly. Nature Communications, 2018. Github.

A.A. Ginart, S. Das, J.K. Harris, R. Wong, H. Yan, M. Krauss, P.A. Cavazos-Rehg. Drugs or Dancing? Using Real-Time Machine Learning to Classify Streamed "Dabbing" Homograph Tweets. IEEE International Conference on Healthcare Informatics (ICHI), 2016.

Education

• PhD in Electrical Engineering, Stanford University (2022) 

• MS in Electrical Engineering, Stanford University (2020)

• BS in Computer Engineering, summa cum laude, Washington University in St. Louis (2017)