Amber (Jiale) Wang: software engineer and ML-for-software-engineering researcher

Tidying up the room

3D is not available on this device

You can still browse: Projects · Research · About · Contact

Projects · 2024–2026

Things I've built.

From TPU performance tooling at Google and LLM-powered code analysis at CGI to research on test evolution and code coverage, most of my work sits where machine learning meets software engineering.

FeaturedGoogle · 2026

An agent that tunes TPU training configs

Built an autonomous agentic configuration optimizer with a reward system and context management. On 4 training workloads it reached up to 16% higher TPU throughput. I also enabled HybridSim, a TPU performance simulator, for third-party workloads, with accuracy gains of up to 90%.

AI agentsTPUPerformance simulation
ResearchAIWare 2026

A live benchmark for test and code co-evolution

An executable, continuously updated benchmark for evaluating how well tools evolve tests alongside the code they cover.

BenchmarksSoftware testingLLMs
More →
IndustryCGI · 2025

LLM-powered codebase insights

An AI code analysis platform with code embeddings, indexing, AST parsing and a RAG pipeline that gives business analysts and developers answers about large codebases.

RAGLLMsAST parsing
ResearchASE 2024

Efficient incremental code coverage

Speeding up code coverage analysis for regression test suites by recomputing only what changed between versions.

Program analysisRegression testing
More →

Research · University of Waterloo

Machine learning for software.

I'm a Master of Mathematics student in Computer Science at the University of Waterloo, advised by Pengyu Nie. I study how machine learning can improve software engineering, especially test automation, NLP-based code analysis and data-driven development.

Interests

AI infrastructureML for software engineeringTest evolution & maintenanceAI coding agentsLLMs for codeBenchmarking & evaluation

Publications

Off the clock

Life, lately.

I'm Amber Jiale Wang.

A software engineer and researcher. I build ML-powered tools that help developers test, understand and ship code.

Amber Wang
Based inWaterloo & Toronto, Ontario
CurrentlyMMath CS @ Waterloo
Emailjiale.wang@uwaterloo.ca
ElsewhereLinkedIn  ·  GitHub  ·  Google Scholar

Download résumé ↓

Education

University of Waterloo logo

Master of Mathematics, Computer Science · University of Waterloo

Sept 2024 – Dec 2026 · Advisor: Pengyu Nie · GPA 97/100

Queen's University logo

Bachelor of Computing, Mathematics & Analytics · Queen's University

Sept 2019 – Apr 2024 · GPA 4.0/4.3

Experience

Google logo

Software Engineer Intern · Google

May 2026 – Aug 2026 · Mountain View, CA

  • Enabled HybridSim, a TPU performance simulator, for third-party workloads, improving accuracy by up to 90%.
  • Built an autonomous agentic configuration optimizer with a reward system and context management, reaching up to 16% higher TPU throughput on 4 training workloads.
CGI logo

AI Software Developer · CGI

Sept 2025 – Dec 2025 · Toronto, ON

  • Developed an AI-driven code analysis platform with code embedding, indexing, AST parsing and a RAG pipeline, using LLMs to produce codebase insights.
  • Enhanced a Ticket Analyzer with NLP clustering, classification pipelines and a similarity-scoring engine.
RBC logo

Machine Learning Developer · Royal Bank of Canada

May 2025 – Aug 2025 · Toronto, ON

  • Prototyped and evaluated agentic AI coding tools (Cursor, Devin, Windsurf) on real engineering tasks.
  • Built automated Python benchmarks across curated repositories, measuring task success rate and code correctness.
Ericsson logo

Automation Engineer · Ericsson Canada

Sept 2022 – Dec 2023 · Ottawa, ON

  • Built 10+ automated workflows with Power Automate, Azure and Python, producing 20+ daily dashboards and cutting manual work by 90%.
  • Developed a license-request email server serving 400+ users, and a React cause-and-effect chain builder.