Systems research & applied engineering

Hongchen (Steven) Yang

Computer science student at Brandeis University

I study adaptive storage systems and build AI agent workflows.

EmailGitHubRésumé
Hongchen (Steven) Yang portrait
Waltham, MA
Institution
Brandeis University
Degree
B.S. Computer Science
Graduation
Expected December 2026
Academic record
GPA 3.748 / 4.0. Dean's List every completed semester.

Research

Storage systems, changing workloads, and the data behind them.

Primary research

FluidLSM and adaptive RocksDB tuning

Smart & Scalable Data Systems Lab · Brandeis University

May 2025 to present

I study how storage systems respond to changing workloads.

Research question
How do skew, burstiness, and changing access patterns affect compaction and performance in LSM tree systems?
Current work
I build controlled RocksDB benchmarks and group dependent configuration parameters to narrow the tuning search space. I use these benchmarks to evaluate adaptive tuning methods for FluidLSM.

Methods RocksDB · LSM trees · Bayesian optimization · decision tree surrogate models · lightweight online learning

Related study

Data fragmentation and SQL generation from text

Data Science Intern (Independent Study) · Brandeis University

Jan 2026 to May 2026

A controlled benchmark for measuring how fragmented administrative data affects SQL generation from text.

Question
How can we vary data fragmentation consistently when comparing systems that generate SQL from text?
Work
I built synthetic schemas and configurable fragmentation generators. I used Python ETL, provenance tracking, and join and query evaluation to compare the original and fragmented data.

Methods Python · SQL · synthetic data · provenance · SQL generation

Publication

Published paper

From Single-View to Multi-view: Learning Informative Graphs for Robust Subspace Segmentation

Dazhai Yang, Jiao Liu, Hongchen Yang, Hualin Liu, and Tianzhe Lou

Advances in Artificial Intelligence, Electronic Instruments and Information Systems
Springer, 2026, pages 82 to 96.

Third author. Contributed in a supporting role.

View paper

Experience

Building workflows, working with teams, and teaching.

Most recent internship

AI Development Intern

Hefei City Cloud Data Center Co., Ltd.

Jun 2026 to Aug 2026

I worked on email reimbursement automation, agent memory, and retrieval.

Built with Java · Spring Boot · Vue · Qdrant · agent memory · workflow orchestration

View reimbursement workflow code
  • Built a workflow that reads reimbursement emails and attachments. It extracts merchant names, amounts, and dates and sends structured records to the reimbursement system.
  • Developed memory assignment and vector retrieval nodes for persistent agent context, filtered search, and memory lifecycle management.
  • Extended Java services and Vue interfaces for memory resources and workflow nodes. Migrated workflows and integrated database, vector store, and model services.

Additional experience

  1. Jan 2026 to May 2026

    Teaching Assistant, Introduction to Database

    Brandeis University

    Led office hours and review sessions for about 50 students and graded coursework on SQL, data modeling, normalization, indexing, and query optimization.

  2. May 2024 to Aug 2024

    Software Engineering Intern

    Shanghai Development Center of Computer Software Technology

    Traced a workflow that generates SQL from text and worked with a Spring Boot and MyBatis backend. Used Linux tools to run services and inspect logs. Refactored Java service logic.

Public Code

Tools I build and maintain.

Maintained public projects.

  • Python · Rust

    fpstreams

    A Python library for typed, lazy data pipelines. Supports synchronous streams, structured async concurrency, record transforms, and optional Rust execution.

    A small pipeline
    (flow([1, 2, 3])
    .filter(lambda n: n > 1)
    .to_list())

# [2, 3]
    PyPIDocsGitHub
  • Python

    dependency-checker

    A skill and MCP interface for coding agents to inspect dependencies across languages. Indexes declarations, resolved versions, source usage, security results, and policy findings.

    Follow the evidence
    pyproject.toml  →  declarations
uv.lock         →  locked versions
src/            →  source references
    GitHub

Other Work

Privacy research and tools for editing notebooks.

  1. Blacklight privacy detection

    Helped research, develop, and validate detection of TikTok and X tracking pixels in Blacklight. The privacy scanner has processed more than 18 million scans.

    Read the Blacklight update

  2. Jupyter Notebook Git Marks

    A VS Code extension that marks added, modified, and deleted source lines in Jupyter Notebook cells. Updates as you type, including changes you have not saved.

    VS Code MarketplaceGitHub

Technical Profile

Programming
Java · C++ · Python · Go · Rust · JavaScript · TypeScript · SQL
Systems & Data
RocksDB · PostgreSQL · MySQL · DuckDB · Redis · Qdrant
Backend & Web
Spring Boot · MyBatis · Vue · Astro · HTML · CSS
Tooling
Docker · Linux · Git · GitHub Actions
Languages
Mandarin Chinese (Native) · English (Fluent)

Contact

Email me about database research, AI systems, or software engineering.

EmailLinkedInGitHubRésumé