Charith Peris
Charith Peris is a senior applied scientist at Amazon whose work focuses on responsible AI for language models. He studied physics at Northeastern, was selected for a 2015 graduate award, and moved from astrophysics into Amazon Alexa’s natural-language-understanding work. His AI projects include AlexaTM 20B, the MASSIVE multilingual dataset, language-model detoxification, and multi-agent chain-of-thought training. As of October 2026, he leads applied-science teams at Amazon AGI on responsible AI for foundation models and agents, including reinforcement learning from human feedback.
Timeline
-
2015
An experiential-learning award
A physics doctoral candidate at Northeastern, he is selected for its Outstanding Graduate Student Award for Experiential Learning in 2015.
-
2022–2023
Multilingual models and data
In 2022, he coauthors AlexaTM 20B, a 20-billion-parameter multilingual sequence-to-sequence model that achieves state-of-the-art performance on one-shot summarization tasks. In 2023, he coauthors MASSIVE, a dataset of one million labelled virtual-assistant utterances spanning 51 languages and 18 domains.
-
2024
Fine-tuning for detoxification
In 2024, he presents regularized fine-tuning that trains language models to follow output constraints while preserving performance.
-
2025
Multi-agent reasoning data
In 2025, he and Tharindu Kumarage report a multi-agent method for generating and refining chain-of-thought training data, with an average 29% improvement across a battery of benchmarks.
-
By 2026
From astrophysics to Alexa
By 2026, he moves from astrophysics into machine learning through Amazon Alexa’s natural-language-understanding work.
-
February 2026
Nova Responsible AI
In February 2026, he is a senior applied scientist in the Nova Responsible AI team at Amazon AGI.
-
May 2026
Leading RLHF post-training
In May 2026, he leads the reinforcement-learning-from-human-feedback phase of model development.
-
As of October 2026
Responsible AI at Amazon
As of October 2026, he leads applied-science teams at Amazon AGI on responsible AI for foundation models and agents.
Career
- Since 2026AmazonSenior Applied Scientist
Education
- Northeastern UniversityPhD candidate in Physics
Most-cited works
- MASSIVE: A 1M-Example Multilingual Natural Language Understanding Dataset with 51 Typologically-Diverse Languages
- Alexa Teacher Model
- Privacy in the Time of Language ModelsWith Richard Zemel
- AlexaTM 20B: Few-Shot Learning Using a Large-Scale Multilingual Seq2Seq Model
- Controlling the Extraction of Memorized Data from Large Language Models via Prompt-Tuning
Sources and credits
Photo: We have no published photo of Charith Peris yet. Send us a free-licensed photo.
- Northeastern announces year-end award winners
- Charith Peris - Amazon Science
- MASSIVE: A 1M-Example Multilingual Natural Language Understanding Dataset with 51 Typologically-Diverse Languages
- Detoxification of large language models via regularized fine-tuning
- Multiagent AI for generating chain-of-thought training data
- Dr. Charith Peris: Navigating AI’s next frontier
- Building trust into AI
- Charith Peris – AI/ML Leader


