Jwala Dhamala

Timeline

1. 2014–2020

   PhD research at RIT

   In 2014, she begins a PhD in computing and information sciences at the Rochester Institute of Technology, in Linwei Wang's Computational Biomedicine Laboratory. Her doctoral research develops probabilistic machine learning, including Bayesian optimization, Gaussian processes and generative models, for personalized cardiac simulation.

2. 2018–2019

   MICCAI Young Scientist finalist

   At MICCAI 2018, she presents a method that embeds a variational auto-encoder into Bayesian optimization to estimate tissue excitability in a cardiac electrophysiological model, with more than a tenfold gain in efficiency. The paper is a finalist for the MICCAI Young Scientist Award, as is her Graph BO-VAE paper at MICCAI 2019.

3. December 2019–February 2020

   Research scientist at Alexa AI

   She joins Amazon's Alexa NU-AI group as a research scientist in December 2019. In February 2020, she defends her RIT dissertation on Bayesian active learning for personalization and uncertainty quantification in cardiac electrophysiological models.

4. 2021

   The BOLD bias benchmark

   In 2021, she and six coauthors introduce BOLD, a dataset of 23,679 English text generation prompts that measures social bias across profession, gender, race, religion and political ideology, together with new metrics for toxicity, psycholinguistic norms and gender polarity. Presented at the ACM FAccT conference in 2021, the study finds that most of three popular language models show more social bias than human-written Wikipedia text.

5. 2021–2022

   Co-organizing TrustNLP

   She co-organizes the TrustNLP Workshop at NAACL in May 2021 and May 2022, and the Responsible AI Workshop at KDD in July 2021. In April 2022, she coauthors an Amazon Science overview of trustworthy machine learning at Alexa AI, covering privacy-preserving machine learning, federated learning and bias mitigation. In May 2022, three of her papers are accepted at ACL 2022 and its Findings.

6. 2023–2024

   JAB and Tree-of-Traversals

   In 2023, she coauthors a FAccT paper on centering transgender and non-binary voices to measure biases in open language generation, and JAB, a framework that probes and improves the robustness of black-box models through adversarial prompting and belief augmentation. At ACL 2024, she and her coauthors present Tree-of-Traversals, a zero-shot reasoning algorithm that augments black-box language models with knowledge graphs.

7. 2025–2026

   Responsible AI evaluation for Nova

   She leads Responsible AI quantitative evaluation for Amazon Nova 1.0 and Nova 2.0, and the Amazon Nova technical report appears in 2025. With Sharon Li and other coauthors, she develops LH-Deception, a multi-agent simulation framework that measures deception by language models over long sequences of interdependent tasks; the paper is accepted at ICLR 2026.

8. As of October 2026

   Senior scientist at Amazon AGI

   She is a senior applied scientist in Amazon's Artificial General Intelligence organization, where she works on responsible-AI evaluations of the company's models. In May 2026, she explains that the team evaluates models from pretraining through predeployment and is working to automate more of the evaluation process. In July 2026, she co-organizes the sixth TrustNLP Workshop at ACL 2026 in San Diego as a member of Amazon Nova RAI.