Marvin Gao
Founder & Hands On Agent System Builder at Parallight Inc
Foster City, California, United States
Summary
🤩
Rockstar🎓
Top SchoolMarvin Gao is a founder and AI leader with 11 years of experience building production-ready ML systems and developer platforms, currently scaling KeploreAI to automate infrastructure, config, and compute orchestration for AI R&D. He combines deep research experience in multi-agent reinforcement learning (MARL)—including implementing HAPPO in the MARLlib ecosystem—with hands-on engineering across backend systems and NLP toolchains from roles at IBM, Alibaba, and Ant Financial. Marvin has led both academic projects (Johns Hopkins) and large product teams, and previously built and exited an ML education platform that was acquired by a unicorn group where he served as VP. He brings a rare blend of startup founder grit and enterprise delivery, having shipped mission-critical AI solutions for banks, logistics, and energy. Based in California, he focuses on RL and multi-agent intelligence while also prioritizing developer experience and reproducibility in research workflows. A less obvious strength is his track record of turning research prototypes into operational systems that reduce cost and manual work at national scale.
11 years of coding experience
7 years of employment as a software developer
Johns Hopkins UniversityComputer Science & Software Engineering, Computer Science & Software Engineering at Zhejiang University
English, Chinese
Github Skills (9)
text-manipulation10
deep-reinforcement-learning10
pytorch10
python10
reinforcement-learning10
text-processing10
natural-language-processing9
ray9
nlp9
Programming languages (9)
Github contributions (5)
computing-intelligence/AI_courses_jupyters_and_slides
Apr 2018 - Dec 2022
Jupyter notebooks and slides for the AI / NLP / text-mining courses (2018–2021)
Role in this project:
Backend Developer
Backend DeveloperContributions:211 commits, 8 PRs, 170 pushes in 4 years 8 months
Contributions summary:Marvin's commit demonstrates the implementation of a basic Python-based edit distance calculation, likely for text processing tasks. The code involves importing the `editdistance` and `jieba` libraries for calculating edit distances and segmenting Chinese text. The user defines functions to perform these operations and applies them to example strings to assess the similarity between them, demonstrating a focus on core functionality within a text-based problem.
aijupyter-notebooknatural-language-processingtext-mining
Replicable-MARL/MARLlib
Mar 2022 - Jan 2023
One repository is all that is necessary for Multi-agent Reinforcement Learning (MARL)
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:201 commits, 10 PRs, 103 pushes in 9 months
Contributions summary:Marvin implemented the HAPPO (Heterogeneous-Agent Proximal Policy Optimization) algorithm within the MARLlib framework. Their contributions include the creation of utilities and core components for HAPPO. These changes demonstrate an understanding of reinforcement learning algorithms, specifically PPO, and multi-agent reinforcement learning (MARL) concepts.
multi-agent-reinforcement-learningrayrllibpytorchdeep-reinforcement-learning
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.