Seyi Feyisetan

I am a Principal Scientist at Amazon working on LLMs. Prior to that, I led research at Meta on DP and MPC. I previously sat on the research advisory board of the IAPP and I am the lead organizer of PrivateNLP workshop series (WSDM2020, EMNLP2020, NAACL2021, NAACL2022, ACL2024). I have been granted 4 patents and I have a PhD in Computer Science from the University of Southampton.

seyi dot feyisetan at gmail

LinkedIn: in/SeyiFeyisetan

Full list on Google Scholar

• Nov 2023: Excited to have written the foreword for Ken Huang's upcoming book on LLM Security!
• Nov 2023: Reviewer for German National Research Center for Applied Cybersecurity (ATHENE)
• Oct 2023: PrivateNLP accepted as a workshop in ACL 2024
• Aug 2023: New patent on Information Uniqueness Determination [pdf]
• Jul 2023: TrustNLP at ACL panel with Hal Daumé III and others
• May 2023: Tutorial on Privacy Preserving NLP at EACL 2023
• Apr 2023: Invited talk at Carnegie Mellon University
• Mar 2023: Invited talk Amazon Data Conference
• Jan 2023: End of tenure on research advisory board of the IAPP

• Dec 2022: Invited talk at the University of Maine
• Dec 2022: New paper at SDM 2023 TEM: High Utility Metric Differential Privacy on Text. [pdf]
• Dec 2022: New patent Calibrated Noise for Text Modification [pdf]
• Nov 2022: Invited talk at PECB Conference (wasn't able to make it)
• Oct 2022: Keynote for 3rd AMLC Privacy Preserving Machine Learning workshop (internal Amazon)
• Sep 2022: Back at Amazon as a Principal Applied Scientist, working on Trust and Privacy
• Jul 2022: Lead organizer for 4th PrivateNLP workshop at NAACL 2021 [website]
• Jun 2022: Leading a session at the joint DP workshop hosted by Google and Meta [website]
• May 2022: New intern - working with Yangsibo Huang over the summer on privacy! 
• Apr 2022: New patent Data-preserving text redaction for text utterance data [pdf]
• Jan 2022: New paper at AISTATS 2022 Reconstructing Test Labels from Noisy Loss Functions [pdf]

• Dec 2021: PriML@ NeurIPS panel with Helen Nissenbaum, Aaron Roth and ‪Christine Task [video]
• Nov 2021: Invited talk at the University of Maine
• Oct 2021: New paper at PriML@ NeurIPS Reconstructing Test Labels from Noisy Loss Scores. [pdf]
• Aug 2021: Started at Facebook!
• Jul 2021: New paper at ICML TPDP TEM: High Utility Metric Differential Privacy on Text. [pdf]
• Jul 2021: New paper at ICML ML4DATA Preserving Privacy of Text Data Efficiently on Device. [pdf]
• Jun 2021: Lead organizer for 3rd PrivateNLP workshop at NAACL 2021 [website]
• Jun 2021: New patent Privacy and intent-preserving redaction for text utterance data [pdf]
• Jun 2021: Area Chair for Personalization and Recommendation at AMLC 2021 (internal Amazon)
• May 2021: Co-organizer - 2nd AMLC Privacy Preserving Machine Learning workshop (internal Amazon)
• May 2021: New paper at ICML 2021 Label Inference Attacks from Log-Loss Scores. [pdf]
• May 2021: Best paper award at FLAIRS 2021 [pdf]
• May 2021: Senior Program Committee member at CIKM 2021
• Apr 2021: Invent and Simplify award (internal Amazon)
• Apr 2021: New paper at PrivateNLP@ NAACL-HLT 2021 On a Utilitarian Approach to Privacy-Preserving Text Generation. [pdf]
• Apr 2021: New paper at TrustNLP@ NAACL-HLT 2021 Private Release of Text Embedding Vectors. [pdf]
• Feb 2021: Feature on Amazon Science for IAPP board seat [website]
• Jan 2021: Feature on Amazon blog for International Privacy Day [website]
• Jan 2021: Research Advisory Board Member at the International Association of Privacy Professionals (IAPP) [website]

• Nov 2020: Invited talk at the University of Oxford
• Nov 2020: Invited talk at the University of Maine
• Nov 2020: Lead organizer for 2nd PrivateNLP workshop at EMNLP 2020 [website]
• Sep 2020: Co-organizer - 1st AMLC Privacy Preserving Machine Learning workshop (internal Amazon)
• Sep 2020: Best paper award at AMLC Shareable NLP Technologies workshop (internal Amazon)
• Sep 2020: New paper at FLAIRS 2021 Differentially Private Adversarial Robustness Through Randomized Perturbations. [pdf]
• Sep 2020: New paper at FLAIRS 2021 Research Challenges in Designing Differentially Private Text Generation Mechanisms. [pdf]
• Sep 2020: New paper at PrivateNLP@ EMNLP 2020 On Primes, Log-Loss Scores and (No) Privacy. [pdf]
• Sep 2020: New paper at PrivateNLP@ EMNLP 2020 A Differentially Private Text Perturbation Method Using a Regularized Mahalanobis Metric. [pdf]
• Jul 2020: Senior Program Committee member at HCOMP 2020
• May 2020: 2 new patents filed with USPTO
• Feb 2020: Lead organizer for 1st PrivateNLP workshop at WSDM 2020 [website]