ABAB News Logo
Back to Flash News

Anthropic co-founder and CEO Dario Amodei’s Baidu-era experience is made public, including Deep Speech 2 and a quarterly star employee award

LinkedIn experience for Anthropic co-founder and CEO Dario Amodei shows that he worked as a Research Scientist at Baidu Silicon Valley from November 2014 to October 2015. The description says that he worked with Andrew Ng and the team on deep learning and AI research, including speech recognition and natural language processing, and was responsible for most of the machine learning research work on the Deep Speech 2 project.

The Deep Speech 2 paper (arXiv:1512.02595), co-authored by Amodei and others, proposed an end-to-end deep learning method for English and Mandarin speech recognition. The project was later listed by MIT Technology Review as one of the 10 breakthrough technologies of 2016. The LinkedIn description also lists results including relative reductions in word error rate for English and Chinese speech systems, a 10x speedup in the decoder, a 27% increase in Chinese training speed, and a 50% reduction in memory use, as well as the co-development with Jim Fan and Jesse Engel of Baidu’s internal neural network library.

The description also records that he was named a “Star Employee of the Quarter” for the second quarter of 2015. Amodei later left Baidu to join Google Brain, then moved to OpenAI, and co-founded Anthropic in 2021.

Original sources:

  1. Title: This is the longest section of Dario Amodei's work experience on LinkedIn by far. Publisher: Steve Hou (@stevehou) on X Publication date: 2026-10-25 URL：https://x.com/stevehou/status/2114551467830865945 Source nature: social media original post (including LinkedIn screenshot) Verification status: actually read the post and screenshot content in this review, verification date 2026-10-26 Supported content: supports that the LinkedIn Baidu experience is the longest paragraph and includes the Q2 2015 Star Employee award record

  2. Title: Deep Speech 2: End-to-End Speech Recognition in English and Mandarin Publisher: arXiv Publication date: 2015-12 URL：https://arxiv.org/abs/1512.02595 Source nature: academic paper Verification status: actually read the abstract and author list in this review, verification date 2026-10-26 Supported content: supports Amodei as a co-author, the end-to-end method for English and Mandarin recognition, and the paper’s authorship including Jim Fan (Linxi Fan)

  3. Title: Baidu’s Deep-Learning System Rivals People at Speech Recognition Publisher: MIT Technology Review Publication date: 2015-12-16 URL：https://www.technologyreview.com/2015/12/16/9846/baidus-deep-learning-system-rivals-people-at-speech-recognition/ Source nature: media report Verification status: actually read the full text in this review, verification date 2026-10-26 Supported content: supports that Deep Speech 2 drew attention, and the system was close to or better than human performance on some benchmarks

ABAB AI Insight

Dario Amodei’s experience in Baidu’s Silicon Valley AI lab centered on the Deep Speech 2 project. The project used end-to-end neural networks to replace traditional hand-engineered features and multi-module pipelines, handling both English and Mandarin, two quite different languages, and in some tests coming close to or exceeding human transcription performance. The paper and subsequent reporting show that the team achieved performance gains through large-scale data and compute, and Amodei’s LinkedIn description focuses his personal contribution on model architecture and efficiency improvements.

Jim Fan (listed as Linxi Fan in the paper) and Jesse Engel were also Deep Speech 2 authors, and LinkedIn lists the three of them as co-developers of Baidu’s internal neural network library. This early collaboration took place before Amodei joined OpenAI and also before he co-founded Anthropic with several former OpenAI colleagues.

Amodei’s Baidu experience is viewed by some observers as one of the most detailed sections of his career, and it includes the 2015 second-quarter Star Employee award. This experience provided direct speech and deep-learning practice that later underpinned his work leading GPT-series research at OpenAI and founding Anthropic. The observed indicators include the correspondence between the public paper author list and the LinkedIn self-description, as well as the technical continuity of the related project in later AI systems.

ABAB News · Cognitive Law

  1. LinkedIn self-description and the arXiv paper author list jointly confirm Amodei’s involvement in the core work of Deep Speech 2.

  2. The end-to-end speech system covers both English and Mandarin, showing an early attempt at multilingual generalization.

  3. The 2015 Star Employee award appears in the public résumé, providing a traceable record for later career milestones.