Ahmad Beirami

Ahmad Beirami is a research scientist at Google DeepMind, leading new research initiatives on post-training within Gen AI Unit. At Google Research, he led a research team on building safe, helpful, and scalable generative language models. At Meta AI, he led research to power the next generation of virtual digital assistants with AR/VR capabilities through robust generative language modeling. At Electronic Arts, he led the AI agent research program for automated playtesting of video games and cooperative reinforcement learning. Before moving to industry, he held a joint postdoctoral fellow position at Harvard & MIT, focused on problems in the intersection of core machine learning and information theory. He is the recipient of the 2015 Sigma Xi Best PhD Thesis Award from Georgia Tech.

• Area Chair for ICML 2025 (November 2024).

• Area Chair for FAccT 2025 (November 2024).

• Reuse Your Rewards: Reward Model Transfer for Zero-Shot Cross-Lingual Alignment will be presented at EMNLP 2024 (November 2024).

• Invited talk at New York University (November 2024).

• Invited Fall Fourier Talk at University of Maryland (October 2024).

• ECE colloquium talk at University of Minnesota (October 2024).

• Attending 1st COLM (October 2024):

  • main conference: Helping or Herding? Reward Model Ensembles Mitigate but do not Eliminate Reward Hacking

• Senior Area Chair for AISTATS 2025 (September 2024).

• Guest editor for Special Issue on Generative Models at IEEE BITS Magazine (September 2024).

• Invited talk at Emerging Generalization Settings Workshop at Simons Institute (September 2024).

• Invited talk at CRL Annual Symposium (August 2024).

• Invited Talk (recording) at Open AGI Symposium (August 2024).

• Area Chair for ICLR 2025 (August 2024).

• Senior Area Chair (Generation) for ACL 2025 (July 2024).

• Attending ICML 2024 (July 2024):

  • main conference: Controlled decoding from language models.

  • main conference: FRAPPÉ: A group fairness framework for post-processing everything.

  • ES-FoMo-II workshop: Block Verification Accelerates Speculative Decoding.

  • MHFAIA workshop: Reuse Your Rewards: Reward Model Transfer for Zero-Shot Cross-Lingual Alignment.

  • Co-organizer for ICML 2024 Workshop on Next Generation of AI Safety.

  • Co-organizer for ICML 2024 Workshop on Theoretical Foundations of Foundation Models.

• Attending ISIT 2024 (July 2024):

  • main conference: Asymptotics of language model alignment.

  • Tutorial on Language Model Inference: Theory and Algorithms.

• Invited speaker at Workshop on Trustworthy Natural Language Processing at NAACL 2024 (June 2024).

• Enhancing Group Fairness in Online Settings Using Oblique Decision Forests presented at ICLR 2024 as spotlight (May 2024).

• Improving Robustness via Tilted Exponential Layer: A Communication-Theoretic Perspective presented at AISTATS 2024 (May 2024).

• Senior Area Chair for NeurIPS 2024 (March 2024).

• Attended EACL 2024 (March 2024):

  • main conference: Gradient-Based Language Model Red Teaming, also featured in Hugging Face Daily Papers .

  • Invited speaker at Personalization of Generative AI Workshop at EACL 2024.

• Attended AAAI 2024 (February 2024):

  • Invited speaker at Human-Centric Representation Learning Workshop at AAAI 2024.

  • Invited speaker at Synergy of Reinforcement Learning and Large Language Models (RL+LLMs) Workshop at AAAI 2024.

  • Panelist at Synergy of Reinforcement Learning and Large Language Models (RL+LLMs) Workshop at AAAI 2024.

• Invited talk at ITA 2024 on asymptotics of language model alignment (February 2024).

• Invited talk at USC on theory & practice of language model alignment (February 2024).

• TPC member for ISIT 2024 (January 2024).

• Area Chair for the 1st COLM (January 2024).

• New preprint on arXiv, Theoretical guarantees on the best-of-n alignment policy with a thread describing the results (January 2024).

• Invited speaker at Human-Centric Representation Learning Workshop at AAAI 2024 (December 2023).

• Break it, Imitate it, Fix it: Robustness by Generating Human-Like Attacks accepted to TMLR (December 2023).

• New preprint on arXiv, Helping or Herding? Reward Model Ensembles Mitigate but do not Eliminate Reward Hacking and also featured in Hugging Face Daily Papers (December 2023).

• Attended NeurIPS 2023 (December 2023):

  • main conference: SpecTr: Fast Speculative Decoding via Optimal Transport

  • main conference (Spotlight): Uncovering the Hidden Dynamics of Video Self-supervised Learning under Distribution Shifts 

  • DistShift workshop: Reward Model Underspecification in Language Model Alignment

  • OTML workshop: SpecTr++: Improved transport plans for speculative decoding of large language models

  • SoLaR workshop (Spotlight): Controlled decoding from language models

  • Co-organized the R0-FoMo Workshop

  • Offered a mentorship session on choosing a career path

  • Recognized as an Outstanding AC.

• Improving Diversity of Demographic Representation in Large Language Models via Collective-Critiques and Self-Voting presented at EMNLP 2023 (December 2023).

• Area Chair for FAccT 2024 (November 2023).

• Invited speaker at the FDS Workshop: "Theory and Practice of Foundation Models" at Yale University (October 2023).

• Panelist at Robotics and AI Symposium 2023 in Queen's University (October 2023).

• New preprint on arXiv, Controlled Decoding from Language Models and also featured in Hugging Face Daily Papers (October 2023).

• Area Chair for ICLR 2024 (August 2023).

• Winning is not everything: enhancing game development with intelligent agents was recognized as the 2023 Outstanding Paper of the IEEE Transactions on Games by IEEE Computational Intelligence Society at IEEE CoG (August 2023).

• Senior Program Committee member for AAAI 2024 Special Track on Safe, Robust and Responsible AI (August 2023).

• New preprint on arXiv, A systematic survey of prompt engineering on vision-language foundation models (July 2023).

• Co-organizer of R0-FoMo:Robustness of Few-shot and Zero-shot Learning in Foundation Models at NeurIPS 2023 (July 2023).

• Action Editor for TMLR (July 2023).

• Invited speaker at Information-Theoretic Methods for Trustworthy Machine Learning workshop at Simons Institute (May 2023).

• On tilted losses in machine learning: Theory and applications  accepted to JMLR (May 2023).

• Co-organizer of ICML 2023 Workshop on "What’s left to TEACH (Trustworthy, Enhanced, Adaptable, Capable and Human-centric) chatbots?" (March 2023).

• Area Chair for NeurIPS 2023 (March 2023).

• Co-organizer of 4 invited sessions on "Differential Privacy" and "Fair & Robust Learning" at ITA 2023 (February 2023).

• Co-organizer of Workshop on Representation Learning for Responsible Human-Centric AI (R2HCAI) at AAAI 2023 (February 2023).

• Robustness through data augmentation loss consistency published at TMLR (January 2023).

• Know thy strengths: Comprehensive dialogue state tracking diagnostics was presented at EMNLP 2022 Findings (December 2022).

• A stochastic optimization framework for fair risk minimization published at TMLR (December 2022).

• Associate Editor for "Signal Processing & Source Coding" and "Machine Learning & Statistics" at IEEE Transactions on Information Theory (November 2022).

• A stochastic optimization framework for fair risk minimization accepted for a contributed talk at TSRML @ NeurIPS 2022 (October 2022).

• Robust conversational agents against imperceptible toxicity triggers presented at NAACL 2022 (July 2022) 

• Database search results disambiguation for task-oriented dialog systems presented at NAACL 2022 (July 2022).

• Co-organized two special sessions on machine learning robustness at ITA 2022 (May 2022).

• Started as a Research Scientist at Google Research (March 2022).

• Co-organized IEEE Internet of Things Magazine Special Issue on "An end-to-end machine learning perspective on Industrial IoT" (March 2022).

• Co-organized  AAAI 2022 Workshop on Human-Centric Self-Supervised Learning (HC-SSL @ AAAI-22) (March 2022).

• New talk, Methods for promoting fairness in deep learning (February 2022).

• New talk, Methods for reliable & responsible neural conversational AI (December 2021).

• New preprint on arXiv, Information-theoretic Bayes risk lower bounds for realizable models (November 2021).

• New preprint on arXiv, FeO2: Federated learning with opt-out differential privacy (October 2021).

• Annotation inconsistency and entity bias in MultiWOZ presented at SIGDIAL 2021 (May 2021).

• An analysis of state-of-the-art models for situated interactive multimodal conversations (SIMMC) presented at SIGDIAL 2021 (May 2021).

• Ditto: Fair and robust federated learning through personalization presented at ICML 2021 (July 2021).

• FERMI: Fair empirical risk minimization via exponential Rényi mutual information  presented at ICML-21 Workshop on Socially Responsible Machine Learning as oral presentation (July 2021).

• 26 contributed papers will be presented at ICML-21 Workshop on Information-Theoretic Methods for Rigorous, Responsible, and Reliable Machine Learning (ITR3 @ ICML-21), see the workshop schedule and the accepted papers (July 2021).

• DAIR: Data augmented invariant regularization presented at ICML-21 Workshop on Uncertainly and Robustness in Deep Learning (July 2021).

• Coded machine unlearning published at IEEE Access (June 2021).

• Training data for DSTC10 track: situated interactive multimodal conversations (SIMMC) 2.0 was released  (June 2021).

• Ditto: Fair and robust federated learning through personalization received best paper award from SECML @ ICLR-21 (May 2021).

• Co-organized ICLR 2021 Workshop on Responsible AI (RAI @ ICLR-21) (May 2021).

• Co-organized ICLR 2021 Workshop on Neural Conversational AI: Bridging the gap between research and real world (NeuCAIR @ ICLR-21) (May 2021).

• DVD: A diagnostic dataset for multi-step reasoning in video grounded dialogue accepted to ACL 2021 (May 2021).

• Tilted empirical risk minimization (TERM) presented at ICLR 2021 (May 2021). 

• Offered a mentorship session for PhD students/postdocs at AISTATS 2021 (April 2021).

• Co-organizing ICML 2021 Workshop on Information-Theoretic Methods for Rigorous, Responsible, and Reliable Machine Learning (ITR3 @ ICML-21) (April 2021).

• Video presentation for tilted empirical risk minimization (TERM) online (March 2021).

• Panelist in "Machine Learning for Internet of Things" panel @ IEEE CCNC 2021 (January 2021).

• A fluorescence sandwich immunoassay for the real-time continuous detection of glucose and insulin in live animals published at Nature Biomedical Engineering (December 2020).

• Situated and interactive multimodal conversations (SIMMC) presented at COLING 2020 (December 2020).

• Resource constrained dialog policy learning via differentiable inductive logic programming (DILOG) presented at COLING 2020 (December 2020).

• Multiple PhD student internships at Facebook Conversational AI Research (CAIR) Team for Winter/Summer/Fall 2021 (October 2020). 

• DSTC9 Track4 "Situated and interactive multimodal conversations (SIMMC)" winners announced (October 2020). 

• Competitive balance in team sports games presented at CoG 2020 (August 2020).

• Rényi fair inference (RFI) presented at ICLR 2020 (August 2020).

• Fair resource allocation in federated learning (q-FFL) presented at ICLR 2020 (August 2020).

• Facebook AI blogpost SIMMC: A data set for developing next-generation shopping assistants online (August 2020). 

• Winning is not everything: enhancing game development with intelligent agents published at IEEE Transactions on Games (June 2020).

• Centralized vs decentralized targeted brute-force attacks: guessing with side-information published at IEEE Transactions on Information Forensics and Security (June 2020).

• Co-organized 2020 Facebook Conversational AI Summit (May 2020).

• Simple team sports simulator (STS2), an environment for multi-agent learning in team sports games, open sourced (May 2020).

• Expert Reviewer Certificate, Transactions on Machine Learning Research (2023).

• Outstanding Paper Award, IEEE Computational Intelligence Society (2022).

• Best Paper Award, ICLR-21 Workshop on Security and Safety in Machine Learning Systems (2021).

• Outstanding Reviewer, ICML, NeurIPS, & ICLR (2020-2022)

• Distinction in Teaching Award, Harvard University (2017)

• Exemplary Reviewer, IEEE Transactions on Communications (2015)

• 2015 Sigma Xi Best Ph.D. Thesis Award, Georgia Tech (2015)

• 2013-2014 Graduate Research Assistant Excellence Award, School of ECE, Georgia Tech (2014)

• Outstanding Research Award, Center for Signal and Information Processing, Georgia Tech (2014)

• Outstanding Service Award, Center for Signal and Information Processing, Georgia Tech (2014)

• Best Student Paper Nomination, 51st IEEE International Midwest Symposium on Circuits and Systems (2008)

• Top 3 in Class, Department of Electrical Engineering, Sharif University of Technology (2007)

• Bronze Medal, 20th Iranian National Mathematics Olympiad (2002)

Machine Learning (conferences, journals, & workshops)

Preprint

• P. Sarkar, A. Beirami, and A. Etemad, "Uncovering the hidden dynamics of video self-supervised learning under distribution shifts," arXiv preprint 2306.02014. [paper]

• J. Gu, A. Beirami, X. Wang, A. Beutel, P. Torr, Y. Qin, "Towards robust prompts on vision-language models," arXiv preprint 2304.08479. [paper]

• N. Aldaghri, H. Mahdavifar, and A. Beirami, "Federated learning with heterogeneous differential privacy," arXiv preprint 2110.15252. [paper]

  • Also presented at NeurIPS-21 Workshop on New Frontiers in Federated Learning: Privacy, Fairness, Robustness, Personalization and Data Ownership (NFFL @ NeurIPS 2021).

Accepted/Published

• X. Ma, S. Mishra, A. Beirami, A Beutel, and J. Chen, "Let's do a thought experiment: Using counterfactuals to improve moral reasoning," ICML-23 Nueral Conversational AI Workshop. [paper]

• T. Li,* A. Beirami,* M. Sanjabi, and V. Smith, “On tilted losses in machine learning: Theory and applications," Journal of Machine Learning Research (JMLR 2023). *Equal contribution. [paper][video][code]

• A. Lowy,* R. Pavan,* S. Baharlouei,* M. Razaviyayn, and A. Beirami, "A stochastic optimization framework for fair risk minimization," Transactions on Machine Learning Research (TMLR 2022). *Equal contribution. [paper][code]

  • Also presented at ICML-21 Workshop on Socially Responsible Machine Learning (SRML @ ICML 2021). *Equal contribution. Oral presentation. 

• H. Cho, C. Sankar, C. Lin, K. Ram Sadagopan, S. Shayandeh, A. Celikyilmaz, J. May, and A. Beirami, "Know thy strengths: Comprehensive dialogue state tracking diagnostics," 2022 Conference on Empirical Methods in Natural Language Processing (EMNLP 2022). [paper]

  • Also presented at ICLR-22 Workshop on ML Evaluation Standards (SMILES @ ICLR 2022).

• T. Huang, S. Halbe, C. Sankar, P. Amini, S. Kottur, A. Geramifard, M. Razaviyayn, and A. Beirami, "DAIR: Data augmented invariant regularization," Transactions on Machine Learning Research (TMLR 2022). [paper][code]

  • Also presented at ICML-21 Workshop on Uncertainty & Robustness in Deep Learning (UDL @ ICML 2021). 

• N. Mehrabi, A. Beirami, F. Morstatter, and A. Galstyan, "Robust conversational agents against imperceptible toxicity triggers," 2022 Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL 2022). [paper]

• K. Qian, S. Kottur, A. Beirami, S. Shayandeh, P. Crook, A. Geramifard, Z. Yu, and C. Sankar, "Database search results disambiguation for task-oriented dialog systems," 2022 Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL 2022). [paper]

• H. Le, C. Sankar, S. Moon, A. Beirami, A. Geramifard, and S. Kottur, "DVD: A diagnostic dataset for multi-step reasoning in video grounded dialogue," The 59th Annual Meeting of the Association for Computational Linguistics, August 2021 (ACL 2021). [paper][code]

• K. Qian, A. Beirami, Z. Lin, A. De, A. Geramifard, Z. You, and C. Sankar, "Annotation inconsistency and entity bias in MultiWOZ,"  The 22nd Annual Meeting of the Special Interest Group on Discourse and Dialogue, July 2021 (SIGDIAL 2021). [paper][video]

• S. Kottur,* P. A. Crook,* S. Moon,* A. Beirami, E. Cho, R. Subba, and A. Geramifard, "An analysis of state-of-the-art models for situated interactive multimodal conversations (SIMMC)," The 22nd Annual Meeting of the Special Interest Group on Discourse and Dialogue, July 2021 (SIGDIAL 2021). *Equal contribution. [paper][code][video]

  • Also presented at AAAI-21 Workshop on Ninth Dialog System Technology Challenge (DSTC9 @ AAAI 2021). *Equal contribution. Oral presentation.[paper][slides]

• T. Li, S. Hu, A. Beirami, and V. Smith, "Ditto: Fair and robust federated learning through personalization," The Thirty-eighth International Conference on Machine Learning, July 2021 (ICML 2021). [paper][code][poster][video]

  • Also presented at ICLR-21 Workshop on Security and Safety in Machine Learning Systems (SECML @ ICLR 2021). Oral presentation. Best Paper Award. 

  • Also presented at NeurIPS-20 Workshop on Scalability, Privacy, and Security in Federated Learning (SpicyFL @ NeurIPS 2020).

• N. Aldaghri, H. Mahdavifar, and A. Beirami, "Coded machine unlearning," IEEE Access, vol. 9, pp. 88137-88150, June 2021 (Access 2021). [paper][code][video]

  • Also presented at AAAI-21 Workshop Towards Robust, Secure and Efficient Machine Learning (RSEML @ AAAI 2021). Oral presentation. 

• T. Li,* A. Beirami,* M. Sanjabi, and V. Smith, “Tilted empirical risk minimization,” 2021 International Conference on Learning Representations, May 2021 (ICLR 2021). *Equal contribution. [paper][video][code]

  • Also presented at ICML-20 Workshop on Uncertainty & Robustness in Deep Learning, July 2020 (UDL @ ICML 2020) . *Equal contribution. 

• C. Gunasekara et al., "Overview of the ninth dialog system technology challenge: DSTC9," AAAI-21 Workshop on Ninth Dialog System Technology Challenge (DSTC9 @ AAAI 2021). All authors contributed equally. [paper]

• Z. Zhou, A. Beirami, P. A. Crook, P. Shah, R Subba, and A. Geramifard, "Resource constrained dialog policy learning via differentiable inductive logic programming," The 28th International Conference on Computational Linguistics, December 2020 (COLING 2020). [paper]

• S. Moon,* S. Kottur,* P. A. Crook,^ A. De,^ S. Poddar,^ T. Levin, D. Whitney, D. Difranco, A. Beirami, E. Cho, R. Subba, and A. Geramifard, "Situated and interactive multimodal conversations," The 28th International Conference on Computational Linguistics, December 2020 (COLING 2020). *Equal contribution. ^Equal contribution. [paper][code]

• S. Baharlouei, M. Nouiehed, A. Beirami, and M. Razaviyayn, “Rényi fair inference,” 2020 International Conference on Learning Representations, August 2020 (ICLR 2020). [paper][code]

• T. Li, M. Sanjabi, A. Beirami, and V. Smith, “Fair resource allocation in federated learning,” 2020 International Conference on Learning Representations, August 2020 (ICLR 2020). [paper][code]

• S. M. Nikolakaki, O. Dibie, A. Beirami, N. Peterson, N. Aghdaie, and K. Zaman, "Competitive balance in team sports games," IEEE Conference on Games, August 2020 (COG 2020). [paper][video]

• Y. Zhao,* I. Borovikov,* F. Silva,* A Beirami,* J. Rupert, C. Somers, J. Harder, J. Kolen, J. Pinto, R. Pourabolghasem, J. Pestrak et al., “Winning is not everything: enhancing game development with intelligent agents,” IEEE Transactions on Games, vol. 12, no. 2, pp. 199-212, June 2020 (TOG 2020). *Equal contribution. Outstanding Paper Award. [paper][poster][code]

  • Also presented at AAAI-19 Workshop on Reinforcement Learning in Games, January 2019 (RLG @ AAAI 2019). *Equal contribution. 

• Y. Zhao, H. Liu, I. Borovikov, A. Beirami, M. Sanjabi, and K. Zaman, “Multi-theme generative adversarial terrain amplification,” ACM Transactions on Graphics, 38(6):200 , November 2019 (SIGGRAPH Asia 2019). [paper][code]

• I. Borovikov, J. Harder, M. Sadovsky, and A. Beirami, "Towards interactive training of non-player characters in video games," ICML-19 Workshop on Human In the Loop Learning, June 2019 (HILL @ ICML 2019). [paper][code]

• Y. Zhao, I. Borovikov, J. Rupert, C. Somers, and A. Beirami, "On multi-agent learning in team sports games," ICML-19 Workshop on Imitation, Intention, and Interaction, June 2019 (I3 @ ICML 2019). [paper][code]

• S. Shahrampour, A. Beirami, and V. Tarokh, "On data-dependent random features for improved generalization in supervised learning," The Thirty-Second AAAI Conference on Artificial Intelligence, pp. 4026-4033, February 2018 (AAAI 2018). [paper][slides]

• A. Beirami, M. Razaviyayn, S. Shahrampour, and V. Tarokh, "On optimal generalizability in parametric learning," 2017 Advances in Neural Information Processing Systems, pp. 3455-3465, December 2017 (NeurIPS 2017). [paper][poster]

Statistics, Information Theory, Signal Processing, & Others (archival journals & conferences)

• M. Nokleby and A. Beirami, "Information-theoretic Bayes risk lower bounds for realizable models," arXiv preprint 2111.04579. [paper]

• M. Poudineh,* C. L. Maikawa,* E. Y. Ma, J. Pan, D. Mamerow, Y. Han, S. W. Baker, A. Beirami, M. Eisenstein, S. Kim, J. Vučković, E. A. Appel, and H. T. Soh, “A fluorescence sandwich immunoassay for the real-time continuous detection of glucose and insulin in live animals,” Nature Biomedical Engineering, vol. 5, pp. 53-63,  January 2021 (Nature BME 2021). *Equal contribution. [paper][code]

• S. Salamatian, W. Huleihel, A. Beirami, A. Cohen, and M. Médard, "Centralized vs decentralized targeted brute-force attacks: guessing with side-information" IEEE Transactions on Information Forensics and Security, vol. 15, pp. 3749-3759, June 2020 (TIFS 2020). [paper]

• A. Beirami and F. Fekri,  "Network traffic compression with side information," IEEE Access, vol. 8., pp. 90023 - 90034, May 2020 (Access 2020). [paper]

• H. Liu, Y. Zhou, A. Beirami, and D. Baron, "Nonlinear function estimation with empirical Bayes and approximate message passing," 57th Annual Allerton Conference on Communication, Control, and Computing, September 2019 (Allerton 2019). [paper][slides]

• S. Salamatian, W. Huleihel, A. Beirami, A. Cohen, and M. Médard, "Why botnets work: distributed brute-force attacks need no synchronization," IEEE Transactions on Information Forensics and Security, vol. 14, no. 9, pp.  2288-2299, September 2019 (TIFS 2019). [paper]

• S. Salamatian, L. Liu, A. Beirami, and M. Médard, "Mismatched guesswork," 2019 IEEE Information Theory Workshop, August 2019 (ITW 2019). [paper]

• A. Beirami, R. Calderbank, M. Christiansen, K. Duffy, and M. Médard, "A characterization of guesswork on swiftly tilting curves," IEEE Transactions on Information Theory, vol. 65, no. 5, pp. 2850-2871, May 2019 (TIT 2019). [paper][code]

• S. Shahrampour, A. Beirami, and V. Tarokh, "Supervised learning using data-dependent random features with application to seizure detection,"  2018 IEEE Conference on Decision and Control, December 2018 (CDC 2018). [paper]

• P. Farajiparvar, A. Beirami, and M. Nokleby, "Information bottleneck methods for distributed learning," 56th Annual Allerton Conference on Communication, Control, and Computing, October 2018 (Allerton 2018). [paper][slides]

• H. Mahdavifar, A. Beirami, B. Touri, and J. S. Shamma, "Global games with noisy information sharing," IEEE Transactions on Signal and Information Processing over Networks, vol. 4, no. 3, pp. 497-509, September 2018 (TSIPN 2018). [paper]

• A. Rezaee, A. Beirami, A. Makhdoumi, M. Médard, and K. Duffy, “Guesswork subject to a total entropy budget,” 55th Annual Allerton Conference on Communication, Control, and Computing, October 2017 (Allerton 2017). [paper][slides]

• S. Salamatian, A. Beirami, A. Cohen, and M. Médard, "Centralized vs decentralized multi-agent guesswork" 2017 IEEE International Symposium on Information Theory, July 2017 (ISIT 2017). [paper]

For a complete list of publications, including papers prior to 2017, refer to Google Scholar!

Research Scientist @ Google Research (Mar 2022-present)

• • Leading research on core machine learning techniques for Responsible AI, and particularly robustness and fairness.

Research Scientist @ Meta AI (Oct 2019-Jan 2022)

• • Led research on Conversational AI, including multimodal conversations, end-to-end dialog management, user satisfaction modeling, and dialog safety/integrity.

  • Led research on evaluating and understanding the tradeoffs between fairness, robustness, privacy, and performance in machine learning.

  • Co-organized the situated and interactive multimodal conversations (SIMMC) track in The Ninth Dialog System Technology Challenge (DSTC9). [GitHub][blogpost][news]

  • Co-organized the 2020 Facebook (company-wide) Conversational AI Summit.

Research Scientist @ Electronic Arts - Data & AI (Apr 2018-Sep 2019)

• • Led research and development on training believable agents for playtesting and gameplaying AI for computer games. 

  • Performed research on graphical content generation for computer games. 

  • Performed research on competitive balance for matchmaking in team online games.

  • Open sourced simple team sports simulator (STS2), an environment for multi-agent learning in team sports games. [GitHub]

  • Hosted several academic and industry visitors, and also visited university campuses to engage in research and deliver technical/recruiting talks (CMU, Caltech, MIT, Harvard, Duke, BU, Columbia, USC, Georgia Tech & Univ. of Michigan).

Postdoctoral Fellow @ Harvard SEAS (Sep 2016-Mar 2018)

• • Derived a scalable approximate leave-one-out cross-validation with optimal theoretical guarantees for a broad range of learning applications including Gaussian mixture models and neural networks. 

  • Provided a data-driven randomized feature selection method for supervised learning, that achieves an effective tradeoff between generalization and scaling. 

Postdoctoral Associate @ MIT EECS (May 2015-Mar 2018)

• • Derived a geometric characterization on an important element of information theory, called guesswork. Our results provide fundamental bounds on error exponents and compression. 

  • Studied vulnerability of security systems against blind distributed botnet attacks without coordination. 

Postdoctoral Associate @ Duke ECE (May 2014- Aug 2016)

• • Provided an information-theoretic characterization of sample complexity in supervised learning when the data of interest lives close to a low dimensional Riemannian manifold.

  • Investigated the existence of Nash equilibrium in multi-agent games with noisy information sharing. 

Ph.D. in ECE (Minor in Math) @ Georgia Tech (2014)

• Derived information theoretic limits of redundancy elimination from network data using universal prediction and unsupervised learning with application to network data compression. 

• Developed a joint memorization and clustering algorithm, called network compression that uses side information to approach the Shannon limit of compression, with low processing overhead. 

• Experimentally demonstrated that around 35% traffic reduction (20% better than baseline method) is achieved on Internet traffic traces of 30 wireless users with the proposed method.

• Proved that these improvements are largely preserved if the fraction of nodes employing network compression exceeds a certain critical value leading to a fundamental phase transition. 

M.Sc. in ECE @ Georgia Tech (2011)

B.Sc. in EE @ Sharif University of Technology (2007)

• "Methods for promoting fairness in deep learning," Meta AI, Amazon AWS AI (2022) [slides]

• "Methods for reliable & responsible neural conversational AI," ASU, Splunk, Google Research, UCL, Vector Institute (2021-2022) [slides]

• "Learning latent representations within reinforcement learning," JP Morgan AI, Facebook AI, Amazon Alexa AI, Bloomberg, Samsung Research, Bosch AI (2019) [slides]

• "Toward cost-aware machine learning," U. Michigan (2019) [slides]

• "Powering games with Data & AI," MIT, Harvard, CMU, Boston University, Georgia Tech, Duke, Columbia, Caltech, USC, U. Michigan (2018-2019). [slides]

• "On generalizability and scalability in learning systems," Technicolor AI Lab, IBM AI, Borealis AI Lab, Electronic Arts (2017-2018). [slides]

• "The geometry of guesswork," Yale, UT Austin, Texas A&M, Rice, UMass Amherst, Harvard, MIT, CO State University, CU Boulder, Georgia Tech (2015-2016). [slides]

• "Packet-level redundancy elimination: an information theoretic approach," MIT, Samsung (2014). [slides]

• Section instructor for Harvard ES 156 — Signals and Systems (Spring 2017) (Rating: 4.79/5.00)

• Recitation instructor for MIT EECS 6.02 — Intro to EECS II: Digital Communication Systems (Fall 2015) (Rating: 6.00/7.00)

• Principal instructor for Duke ECE 587/STA 563 — Information Theory (Spring 2015) (Rating: 4.86/5.00)

Co-organizer (special issue/conference/workshop/seminar):

• AAAI 2023 Workshop on Representation Learning for Responsible Human-Centric AI (R2HCAI @ AAAI-23)

• IEEE Internet of Things Magazine 2022 Special Issue on "An end-to-end machine learning perspective on Industrial IoT"

• AAAI 2022 Workshop on Human-Centric Self-Supervised Learning (HC-SSL @ AAAI-22)

• ICML 2021 Workshop on Information-Theoretic Methods for Rigorous, Responsible, and Reliable Machine Learning (ITR3 @ ICML-21)

• ICLR 2021 Workshop on Neural Conversational AI (NeuCAIR @ ICLR-21)

• ICLR 2021 Workshop on Responsible AI (RAI @ ICLR-21)

• 2020 Facebook Conversational AI Summit

• 2018-2019 EA Data & AI bi-weekly machine learning seminar series 

• 2015 ACM International Conference on Nanoscale Computing and Communication (NANOCOM) 

• 2013-2014 Georgia Tech Center for Signal and Information Processing (CSIP) weekly seminar series 

External dissertation/thesis committee member: 

• Yanting Ma, “Solving large-scale inverse problems via approximate message passing and optimization,” North Carolina State University (2017) [PhD dissertation]

• Quentin Vaucher, "Compressed federated learning," Aarhus University (2021) [MS thesis]

Associate Editor / Reviewer for journals: 

IEEE Trans. Information Theory, IEEE Trans. Signal Processing, IEEE Trans. Communications, IEEE Trans. Circuits and Systems-II, IEEE Trans. Very Large Scale Integration (VLSI) Systems, IEEE Trans. Molecular, Biological, and Multi-Scale Communications, IEEE Journal on Selected Areas in Communications (JSAC), IEEE Journal of Selected Topics in Signal Processing (JSTSP), IEEE Signal Processing Letters, IEEE Internet of Things Magazine (IoTM), Journal of Optical Society of America-B (JOSA-B), Entropy, Computer Graphics Forum (CG)

Area Chair / Reviewer for conferences : 

Intl. Conference on Learning Representations (ICLR), Neural Information Processing Systems (NeurIPS), Intl. Conference on Machine Learning (ICML), Artificial Intelligence and Statistics (AISTATS), Asian Conference on Machine Learning (ACML), Annual Conference of the Association for Computational Linguistics (ACL), The Conference on Empirical Methods in Natural Language Processing (EMNLP), European Chapter of the Association for Computational Linguistics (EACL), Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL), International Conference on Computational Linguistics (COLING), Annual Meeting of the Special Interest Group on Discourse and Dialogue (SIGDIAL), AAAI Conference on Artificial Intelligence (AAAI), Computer Vision and Pattern Recognition Conference (CVPR), European Association on Computer Graphics (Eurographics), IEEE Intl.  Symp. on Information Theory (ISIT), IEEE Conference on Decision and Control (CDC), IEEE Information Theory Workshop (ITW), IEEE Intl. Conference on Computer Communications (INFOCOM), Data Compression Conference (DCC)