Dongwei Jiang
Applied Scientist II at Amazon
California, United States
Summary
👤
Senior🎓
Top SchoolDongwei Jiang is an applied scientist with 11 years of experience who now builds agentic LLMs, reinforcement learning guides, and systems for self-improvement in California. He blends deep research on LLM reasoning and feedback—work that revealed surprising limits in models’ ability to discriminate and incorporate self-generated corrections—with production ML engineering gained across speech and dialogue products. Prior roles at DiDi, Yuanfudao, Shopee and Yuanfudao show a track record of shipping end-to-end ASR/TTS systems, deploying streaming models at scale, and improving low-resource pipelines that drove measurable product impact. He contributes to prominent open-source speech toolkits like ESPNet and Athena, focusing on deployment, cluster integration, and WFST decoding for reproducible production workflows. At Amazon he researches generalized RL agents and lightweight task-specific guide models that boost zero-shot generalization without retraining core models. Known for bridging rigorous academic research (Johns Hopkins, Edinburgh, UT Austin) with pragmatic, deployable ML systems, he uniquely pairs theorem-proving–inspired reasoning work with large-scale speech engineering.
11 years of coding experience
7 years of employment as a software developer
Exchange student, Computer Science, Exchange student, Computer Science at Massachusetts Institute of TechnologyJohns Hopkins UniversityBachelor's degree, Geographic Information System, Bachelor's degree, Geographic Information System at Peking University
Chinese, Chinese, English
Github Skills (21)
slurm10
scripting10
bash10
automation10
script10
tensorflow10
asr10
sh10
speech-recognition10
automations10
shell10
python9
cloud-infrastructure9
devops9
sequence-to-sequence9
Programming languages (8)
Github contributions (5)
athena-team/athena
Dec 2019 - Jan 2021
an open-source implementation of sequence-to-sequence based speech processing engine
Role in this project:
Back-end Developer & DevOps Engineer
Back-end Developer & DevOps EngineerContributions:52 reviews, 136 commits, 142 PRs in 1 year
Contributions summary:Dongwei primarily addressed bugs and made improvements to the speech processing engine, specifically resolving issues related to Recurrent Neural Network Language Models (RNNLM) and updating benchmark configurations. They also worked on configuration files, run scripts, and made changes to the decoding process and solver. Furthermore, the user contributed to fixing assertions in CPU mode within the solver and addressed horovod import issues, suggesting involvement in model training and optimization. The user also incorporated WFST decoding.
sequenceasrctcspeech-recognitionwfst
espnet/espnet
Mar 2019 - Jun 2023
End-to-End Speech Processing Toolkit
Role in this project:DevOps Engineer
DevOps EngineerContributions:6 reviews, 3 PRs, 13 comments in 4 years 3 months
Contributions summary:Dongwei primarily contributes to the project by adding and modifying scripts related to the build and deployment process, particularly focusing on integration with job scheduling systems like `slurm` and `ssh`. Their changes involve configuring command execution environments and adapting scripts to specific cluster environments, suggesting a focus on automating and managing the execution of the project's tasks. The user also adds recipes related to specific datasets.
speech-recognitionspeech-separationchainerspoken-language-understandingspeech-processing
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.