# Xiaochen Yang

> GenAI MLE @ Google Cloud AI | Gemini

Location: San Francisco, California, United States
Profile: https://flows.cv/xiaochen

## Work Experience
### Software Engineer @ Google
Jan 2022 – Present | Mountain View, CA
2025/01 - Present: 
- E2E Agentic Tool-use benchmarks for Gemini quality hillclimbing 
- Gemini Enterprise agent harness quality research
- Post-training (SFT, RL) on horizontal agent capabilities 

2024/02 - 2024/12:
- (20% project) Scalable synthetic STEM QA auto-generation for Gemini 2.0 Pre-training.

2022/07 - 2024/12:
- Ads customer engagement data infra/platforms @ Google Ads

### Software Engineer Intern @ Activision Blizzard Media
Jan 2021 – Jan 2021 | San Francisco Bay Area
• Joined supply-side platform (SSP) team on Ad Tools platform, enabling ad targeting, ad provider bidding, and ad delivery to games that serve more the 300 million users.

• Boosted resource loading time on Ads Tooling Portal’s listing pages by 500ms through building a server-side preloading pagination component using TypeScript.

• Implemented a generalized Promise loader for asynchronous data handling across all pages on Ads Tooling Portal using React higher-order component.

• Streamlined user segmentation process for ad targeting by building a dynamic visual editor using React with Typescript and MobX state management.

• Created backend methods for uploading user segment configuration JSON to Google Cloud Storage buckets using internal Java framework pFlight.

### Undergraduate Research Assistant at Info Lab @ University of Southern California
Jan 2020 – Jan 2021 | Los Angeles, California, United States
Co-developed CrimeForecaster, a crime forecasting deep learning framework that captures crime events’ spatial and temporal dependencies via Diffusion Convolution and Gated Recurrent Network, using TensorFlow (Python)

Represented physical and semantic proximities of neighborhoods as weighted edges in graphs through doing geographical computations in spatial databases using PostgreSQL and PostGIS

Improved forecasting performance by 10% via feature selection through testing different feature combinations on Machine Learning baselines using scikit-learn

Coauthored paper “CrimeForecaster: Crime Prediction by Exploiting the Geographical Neighborhoods’ Spatiotemporal Dependencies”, published in ECML-PKDD 2020

### Full Stack Engineer at Data Lab @ University of Southern California
Jan 2020 – Jan 2021 | Los Angeles, California, United States
Co-developed ReCOVER, an user-interface web application for showcasing predictions from COVID forecasting model SI-kJapha using React (JavaScript) and React Router. Projects got featured by CDC.

Built and maintained the back-end server for collecting and updating worldwide and nationwide COVID infections/deaths cases everyday using Django (Python). Received above 700 requests per day from users.

Co-developed COVID19-Forcast-Bench, a front-end leaderboard app for collecting predictions from 43 different labs and sorting labs on a table based on mean squared error using React

Implemented and beautified front-end forms, graphs and dynamic heat maps using JavaScript libraries including Ant Design, Nivo, amCharts, Recharts, and DevExtreme Reactive

### Orientation Advisor @ University of Southern California
Jan 2019 – Jan 2020
Provided customer service for 5800 number of students and guests through the office’s 11
programs. Facilitated small group discussions for new students regarding academic success, campus resources, university policies, and other common concerns for first-year students.

### Undergraduate Research Assistant @ ETH Zürich
Jan 2020 – Jan 2021 | Zurich, Switzerland
Optimized performance of a semantic role labeling task by 10% via using pre-trained contextualized word embeddings ELMO and BERT from AllenNLP models.

Run and evaluated the performance of event/entity co-reference and extraction models via calculating various scoring matrices using AllenNLP (PyTorch)

Parsed and cleaned XML format documents from NLP corpora for training/testing input via Xml.dom (document model API) using Python


## Education
### Bachelor of Science - BS in Computer Science
University of Southern California
Jan 2018 – Jan 2022

### Bachelor of Science - BS in Applied Mathematics
University of Southern California
Jan 2019 – Jan 2022

### High School Diploma
Beijing National Day School
Jan 2015 – Jan 2018


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

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