# Juntao Li

> Software Engineer at Google

Location: San Bruno, California, United States
Profile: https://flows.cv/juntao

Computer vision, Reinforcement Learning and Distributed System

## Work Experience
### Software Engineer @ Google
Jan 2020 – Present | San Francisco Bay Area
High performance distributed storage system for Ads

### Masters Student @ Carnegie Mellon University
Jan 2018 – Jan 2020 | Pittsburgh, PA

### Research Assistant @ Carnegie Mellon University
Jan 2019 – Jan 2019 | Greater Pittsburgh Area
Autonomous perception project
Researcher Assistant and Team leader | Team Project: Python, Tensorflow, OpenCV
•Designed an innovative traffic gesture recognition algorithm achieving 89.7% accuracy
•Created a 400GB dataset for training; Built a tool with YOLO and OpenCV increasing 90% labeling speed

### Software Engineer Intern @ Splunk
Jan 2019 – Jan 2019 | San Francisco Bay Area
Software Engineer Intern | Individual Project: JavaScript, React
•Designed and Implemented adding new chart algorithm increasing speed by90%
•Designed and implemented resizing/moving items automatically solving overlaps in real-time
•Created dashboard-definition package contains definition and schema used by 5,000 engineers at Splunk
•Designed and implemented customized source code validation (core feature) used by 80,000 customers
•Reduced dashboard build time by 50% in customer side with validation and auto-completion

### Computer Vision Researcher @ SenseTime 商汤科技
Jan 2018 – Jan 2018 | Beijing City, China
The goal of this task was to detect and inpaint freckles in the face.
• Implemented Deeplab, YOLO and Mobile Net using caffe
• Designed freckle segmentation CNN using caffe framework and achieved 75% mIoU
• Implemented inplanting algorithms using patches

### Research Assistant @ Zhejiang University (ZJU)
Jan 2017 – Jan 2018 | Hangzhou, Zhejiang, China
Artificial Intelligence Institute
Worked on reinforcement learning
OGTL is a two stage learning scheme which can reduce the training time and achieve higher winning rate
• Implemented DQN, AC, A3C and Reinforce algorithms using tensorflow framework
• Designed Oponent Guided Tactic-aware Learning (OGTL) scheme
• OGTL algorithm reduces the training time to 10% compared with DQN
• OGTL algorithm’s wining rate is 21% higher than state of the art algorithms
• Pulication: Juntao Li, Yue Hu, Xi Li, Gang Pan and Mingliang Xu, ”Knowledge-Guided Agent-TacticAware Learning for StarCraft Micromanagement,” IJCAI, 2018

### Software Engineer Intern @ Roboy traffic Technology Co. Ltd.
Jan 2016 – Jan 2016 | Changzhou, Jiangsu, China
The goal of this task was to build and maintain the company management website.
• Deployed Spring, Hibernate, Spring MVC frameworks to build and maintain the management website
• Enabled users to monitor the cars’ location, logging and maintain the user privileges


## Education
### Master's degree in Computer Science
Carnegie Mellon University
Jan 2018 – Jan 2020

### Bachelor of Engineering - BE in Computer Software Engineering
Zhejiang University
Jan 2014 – Jan 2018


## Contact & Social
- LinkedIn: https://linkedin.com/in/janathonl

---
Source: https://flows.cv/juntao
JSON Resume: https://flows.cv/juntao/resume.json
Last updated: 2026-03-22