Hi, I'm Abimael

About me

I specialize in speech and language AI, with a particular focus on the messy, real-world scenarios where most systems break down: child speech, multi-speaker environments, and code-switching.

My background combines three years of developmental psycholinguistics research at Penn (studying how children acquire language) with hands-on ML work at xAI and Meta (evaluating LLMs, ASR systems, and multimodal models). This gave me a rare perspective: I understand how human process speech and how to build systems that need to handle it.

Recent work includes:

• Evaluating Grok 4's conversational capabilities and multimodal features at xAI
• Fine-tuning ECAPA-TDNN for child speech language identification at UW
• ASR quality evaluation and transcription pipeline work at xAI and Meta
• Placing 6th overall in the ACL Shared Task for biomedical summarization
• Building CNN-based automation for child gaze analysis

I'm drawn to problems where linguistic theory meets engineering constraints where understanding why speech systems fail (phonetics, prosody, speaker variation) helps you build systems that actually work.

Looking for roles in speech ML, conversational AI, or audio processing where I can work on the hard problems: real-world audio, low-resource languages, and systems that work for diverse speakers.