Yixin Dong
Parttime Researcher at xAI
Pittsburgh, Pennsylvania, United States
Summary
🤩
Rockstar🎓
Top SchoolYixin Dong is a PhD-level researcher and engineer focused on large language model systems, with eight years of experience spanning research internships and part-time roles at xAI, Databricks, and DeepSeek. They specialize in LLM tool calling, agents, pretraining and inference optimization, and practical engineering such as LoRA-customized speculative decoding and MoE model efficiency. Yixin has contributed to prominent open-source work on mlc-llm, improving memory and attention mechanics and adding grammar parsing integrations that streamline model-building pipelines. Based in Pittsburgh with a strong academic foundation from Carnegie Mellon and top grades from Shanghai Jiao Tong, they bridge rigorous research with production-aware implementation. Colleagues describe them as a hands-on problem solver who surfaces elegant compiler- and systems-level fixes that yield measurable speed and memory gains. Outside code, their Github bio hints at an introspective streak that often guides careful, detail-oriented design choices.
8 years of coding experience
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at Carnegie Mellon UniversityBachelor of Science - BS, Computer Science, 4.03 / 4.3, Bachelor of Science - BS, Computer Science, 4.03 / 4.3 at Shanghai Jiao Tong University
Github Skills (8)
compilation10
python10
json-schema10
compile10
llm10
tvm9
language-model9
machine-learning8
Programming languages (3)
Github contributions (5)
mlc-ai/mlc-llm
May 2023 - Nov 2024
Universal LLM Deployment Engine with ML Compilation
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:22 reviews, 23 PRs, 6 pushes in 1 year 6 months
Contributions summary:Yixin contributed to the memory optimization of the model building process by modifying the Llama attention mechanism and the causal mask generation in the `mlc_llm/relax_model/llama.py` file. They also worked on the implementation of a BNF AST and parser for EBNF grammar, introducing new code for parsing and simplifying grammars. Additionally, the user integrated a JSON grammar into the generation pipeline.
language-modelllmmachine-learning-compilationtvm
Ubospica/tvm-develop
Feb 2023 - May 2024
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Contributions:155 pushes, 56 branches in 1 year 3 months
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.