Abimael Hernandez Jimenez

Speech & NLP Engineer with a foundation in developmental psycholinguistics and large-scale model evaluation.

Currently at Amazon AGI. Formerly xAI and Meta.

About

I specialize in the intersection of linguistic theory and machine learning, with a focus on the "noisy" scenarios where AI often struggles: child speech, code-switching, and multi-speaker environments.

My career bridges developmental psycholinguistics research at Penn (studying how humans acquire language) and industry-leading AI evaluation at xAI and Meta. This background allows me to understand the phonetic and prosodic nuances of speech while building the data and evaluation frameworks necessary to scale modern AGI.

Recent Impact

Amazon AGI Data Services: Engineering data quality, evaluation, and annotation strategies for large-scale language systems.
xAI: Evaluated and refined Grok 4’s conversational and multimodal capabilities; contributed to ASR quality and transcription pipeline development.
ACL 2025: Ranked 6th overall in the BioLaySumm Shared Task for biomedical summarization.
UW Research: Fine-tuned ECAPA-TDNN architectures for child speech language identification.
Meta: Focused on ASR quality evaluation and multimodal model performance benchmarks.
Penn: Developed CNN-based automation for child gaze analysis in developmental studies.

Publications

MIRAGES at BioLaySumm2025: The Impact of Search Terms and Data Curation for Biomedical Lay Summarization.
Pong, B., Chen, J.-H., Jiang, J., Jimenez, A., & Vahadi, M. (2025). BioNLP Workshop (ACL 2025).

Cross-linguistic scope ambiguity: An investigation of English, Spanish, and Mandarin.
Song, Y., Hernandez Jimenez, A., & Scontras, G. (2021). Proceedings of the Linguistic Society of America.

Open to conversations about research and engineering roles in speech and language AI.

Connect with me on LinkedIn