# Xiang Gu (顾乡)

> Reference Data @ Citadel

Location: New York, New York, United States
Profile: https://flows.cv/xianggu

Distributed data systems, sports fan.

## Work Experience
### Software Engineer @ Citadel
Jan 2024 – Present | New York, NY
- Worked on a predicate-level market data indexing platform with day-zero + incremental indexers and a mission-critical gRPC service; deployed across SQL Server (on-prem) and GCP Postgres with multi-region support.

- Led development of Archive read APIs for a bitemporal, versioned data subsystem, enabling current-state and historical access to archived market data and driving migration from legacy systems to a unified next-generation data platform.

- Designed and migrated semantic data quality (DQ) framework to R4B platform, eliminating silent failures and reducing evaluation time for hundreds of SPARQL rules 5x; authored internal best-practices for SPARQL/LMDB optimization and built tooling to support weekly production releases and 24/7 DQ monitoring.

- Built entity-to-entity relationship history indexing system, reducing storage footprint by 80% and enabling advanced historical link analysis previously infeasible under predicate-level indexing.

### Member Of Technical Staff @ Cockroach Labs
Jan 2022 – Jan 2024 | New York, New York, United States
Responsible for database schema objects metadata that are used to support SQL queries in CockroachDB, a scalable, highly available, and distributed SQL database.

Key contributor to the development of CockroachDB's next-gen schema change infrastructure that provides online and transactional schema change experience. This is an industry first effort.

### Graduate Teaching Assistant @ The University of Texas at Austin
Jan 2020 – Jan 2021 | Austin, Texas, United States
CS347: Data Management (Fall 20), CS343: Artificial Intelligence (Spring 20, Spring 21, Fall 21)

### Software Engineer Intern @ Google
Jan 2021 – Jan 2021
Built a Spanner-backed datastore for an internal service that makes ML research artifacts discoverable, reusable, and reproducible through effortless lineage tracking.

This storage backend greatly outperformed the existing approach as benchmark tests showed a 2-3 orders of magnitude improvement in query latency (mean, max, stddev, p90 through p99 percentile) while sustaining 40 QPS.

### Software Engineer Intern @ Amazon
Jan 2020 – Jan 2020 | Seattle, Washington, United States
Automated the process of ensuring data consistency (known as ’backfilling’) for Kindle library series grouping; Reduced operation time for backfilling from an average of 45 minutes to less than 5 seconds.

### Research Engineer Intern @ Tencent
Jan 2019 – Jan 2019 | Shenzhen, Guangdong, China
In Tencent AI Lab, I worked with my supervisor and colleagues on applying AI algorithms on agriculture. We studied and developed batch reinforcement learning algorithms to learn planting strategies to help grow cucumbers via remote greenhouse control.

### Undergraduate Research Assistant @ Shanghai Jiao Tong University
Jan 2019 – Jan 2019 | Shanghai City, China
In the APEX lab, I worked under Prof. Weinan Zhang's supervision on exploring possibilities on applying model-based reinforcement learning algorithm to recommendation systems.

### Undergraduate Research Assistant @ University of Alberta
Jan 2018 – Jan 2019 | Edmonton, Canada Area
In the RLAI lab, I worked with Prof. Richard Sutton on empirically evaluating the on-policy version of a modern off-policy learning method called "Emphatic Temporal Difference" (ETD) Learning. 

We observed possible advantages of applying ETD under on-policy training on several known TD counter-examples as well as one classic reinforcement learning problem (MountainCar).

### Undergraduate Research Assistant @ The University of Texas at Austin
Jan 2018 – Jan 2018 | Austin, Texas Area
Under the supervision of Prof. Peter Stone in the LARG group in the computer science department, I worked with a senior Ph.D. student on off-policy methods in reinforcement learning -- methods that separate the target policy the agent tries to learn with the policy that the agent actually deploys to interact with the environment.

We proposed a new framework for off-policy policy gradient methods -- methods that directly manipulate and optimize the policy.


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

---
Source: https://flows.cv/xianggu
JSON Resume: https://flows.cv/xianggu/resume.json
Last updated: 2026-04-05