Travis Johnson
Watson Junior Software Engineer - AlchemyAPI
Wrightwood, California, United States
Summary
👤
Senior🎓
Top SchoolTravis Johnson is a software engineer with a strong physics foundation (M.S., UCSD) and 11 years of hands-on experience applying computational methods to real-world problems, currently contributing to IBM Watson’s AlchemyAPI. He transitioned from medical imaging research—where he implemented clustering algorithms, Radon transform simulations, and authored peer-reviewed work—to backend development and test automation for high-throughput ML inference projects. At vllm he added sampling controls and reliability fixes, showing fluency with model-serving edge cases and rigorous test coverage. Known as a quick study who moves from prototype to production, he combines MATLAB-era algorithmic rigor with modern engineering practices. Based in Wrightwood, CA, he brings a practical work ethic honed in lab, teaching, and field roles and is actively expanding skills in ITK/VTK, object-oriented design, data science, and machine learning.
11 years of coding experience
2 years of employment as a software developer
University of California San DiegoBachelor’s Degree, Physics, 3.9 GPA, Bachelor’s Degree, Physics, 3.9 GPA at University of California, Irvine
Github Skills (12)
algorithm10
upsampling10
pytorch10
subsampling10
sampling10
inference10
pytest10
sampler10
python10
testing10
debug9
transformers8
Programming languages (4)
Github contributions (5)
vllm-project/vllm
Feb 2024 - Mar 2025
A high-throughput and memory-efficient inference and serving engine for LLMs
Role in this project:
Backend Developer & Test Automation Engineer
Backend Developer & Test Automation EngineerContributions:39 reviews, 41 PRs, 123 comments in 1 year 1 month
Contributions summary:Travis implemented the `min_tokens` sampling parameter feature, enhancing the model's generation control. They also contributed several bug fixes related to speculative decoding, Mistral tokenizer edge cases, and other model-specific issues. Furthermore, the user wrote extensive test code and added test cases to ensure the correctness of sampling and speculative decoding features within the VLLM framework.
amdcudadeepseekgpthpu
tjohnson31415/vllm
Feb 2024 - Apr 2025
A high-throughput and memory-efficient inference and serving engine for LLMs
Contributions:127 pushes, 70 branches in 1 year 1 month
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.