Awni Hannun

Career

Awni Hannun completed a PhD in computer science at Stanford University advised by Andrew Ng, and co-led the Deep Speech projects at Baidu's Silicon Valley AI Lab. Hannun also spent time at Zoom and Facebook AI Research (FAIR), was a research scientist at Apple, and is now a Member of the Technical Staff at Anthropic.¹³

Research

Much of Hannun's published research concerns speech recognition. Hannun is first author of Deep Speech (2014), an end-to-end deep learning speech recognizer that does without hand-designed components for noise and speaker variation, and wrote a 2017 Distill article explaining the connectionist temporal classification (CTC) algorithm used to train such models.⁴⁵⁶

Hannun's other work includes a 2017 convolutional network for detecting heart arrhythmias from single-lead electrocardiograms and papers on privacy in machine learning, including a method that uses Fisher information to measure how much a model leaks about its training data.⁷⁸⁶

Open-source work

Hannun's software page lists MLX, an array framework for Apple silicon, alongside earlier projects: GTN, a library for automatic differentiation with weighted finite-state transducers; the Flashlight C++ machine learning library; the wav2letter++ speech recognition toolkit; the CrypTen library for privacy-preserving machine learning on PyTorch; and a C implementation of the RNN Transducer loss.⁹²

Work and contributions

• MLX

  Co-created MLX at Apple. The README says the MLX software suite was initially developed with equal contribution by Awni Hannun, Jagrit Digani, Angelos Katharopoulos, and Ronan Collobert, and lists the four as authors in its citation entry.¹³²

Timeline

1. Dec 2014Posted Deep Speech, an end-to-end speech recognition system, as first author.⁴
2. Jul 2017Co-authored a convolutional neural network study on arrhythmia detection from electrocardiograms.⁷
3. Nov 27, 2017Published "Sequence Modeling with CTC" in Distill.⁵
4. 2018Completed a PhD thesis in computer science at Stanford University.⁶
5. Feb 2021Posted a paper on measuring data leakage in machine learning models with Fisher information.⁸

Other documented roles

Roles the sources document that are not confirmed as current.

• Research Scientist, Apple¹

Honors and recognition

• Best Paper Award, UAI 2021 · Conference on Uncertainty in Artificial Intelligence, 2021⁶

Where they work

• Anthropic 34 people in the wiki

Related people

• Olivier ChafikAlso at Anthropic
• Rahul PatilAlso at Anthropic
• Azalia MirhoseiniAlso at Anthropic
• Nikhil ThoratAlso at Anthropic
• Daniela AmodeiAlso at Anthropic
• Dmytro DzhulgakovInference and serving · ML frameworks and libraries
• Gavin UbertiInference and serving · ML frameworks and libraries
• Tim DavisML frameworks and libraries · Inference and serving
• Joshua LochnerInference and serving · ML frameworks and libraries
• Ian BuckML frameworks and libraries · Inference and serving