Scholar
Julian Salazar
Google Scholar ID: Nd6DvD4AAAAJ
Google DeepMind
Citations & Impact
Citations
2,510
H-index
12
i10-index
12
Publications
20
Co-authors
0
Publications
Resume (English only)
Academic Achievements
- - Published papers on topics such as speech generation and synthesis, large pretrained language models, speech recognition and self-supervision, NLP for low- to zero-resource languages.
- - Specific projects/papers: SpeechSSM, Spectron, SequenceLayers, Very Attentive Tacotron, Mono-to-Binaural, Masked LM Scoring, Transformers without Tears, Meta-Learning the Difference, Transformer + CTC, Align-Refine, DeCoAR, BERTphone, Attention Failures in ASR, Unsupervised Translation, Prompting with Phonemes, 0-shot E2E SLU, 0-shot X-lingual Evals.
- - Erdős number is 2.
Research Experience
- - Staff Research Scientist at Google DeepMind's Frontier AI unit, working on AI models for generative audio, especially in speech and dialogue.
- - Worked over a year in the former Perception team at Google Research.
- - Senior Applied Scientist at Amazon AWS AI for more than five years, contributing to AI cloud services, automatic speech recognition systems, and fast acoustic architectures.
Education
- Graduated from Harvard University with an honors degree in Mathematics and a secondary in Computer Science.
Background
- Research interests include deep learning for human language processing, specifically natural language processing (NLP). Focused on AI models for generative audio, particularly speech and dialogue. Also interested in pure mathematics.
Miscellany
- From the Philippines and Canada; held summer research fellowships at U. Chicago and Cambridge; worked as a full-time mobile developer (Top Hat); teaching fellow (CS50) and course assistant (Differential Topology, Galois Theory); internships at J.P. Morgan S&T and HubSpot; directed a non-audition show choir.
Co-authors