Skip to content
View axlezhao's full-sized avatar

Block or report axlezhao

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
axlezhao/README.md

Hi, I'm Yize (Axle) Zhao 👋

I'm an M.S. student in Data-Enabled Computational Engineering and Science at Brown University, with a B.S. in Data Science from UC Irvine. I build AI agents and the infrastructure that makes them reliable: tool use over MCP, evaluation, and fault-tolerant pipelines. I also focus on data science and machine learning, especially deep learning, with work ranging from search and ranking to physics-informed neural networks.

  • Most recently an AI Agent Developer Intern at Tencent Cloud, where I built agent diagnosis, evaluation and regression-testing infrastructure
  • Interested in LLM agents, evaluation, retrieval and ranking, and physics-informed ML

Featured projects

Project What it is Stack
Wanderly: AI Travel Agent A ReAct agent that plans trips by calling real travel data through an MCP tool server. Works with Claude, DeepSeek or any OpenAI-compatible model, with a fault-tolerant harness and an eval suite Node.js · Go · MCP
AlphaLens An evidence-first equity research workbench that turns SEC filings into cited, point-in-time facts, with automated verification and owner approval before anything is published (open-source beta) TypeScript · Cloudflare Workers · D1
A2Z Search A from-scratch search engine over 51K UCI webpages: SPIMI inverted index, compressed postings, BM25 ranking, 33.8 ms median query latency Python
Learning Dynamical Systems from Data FNN vs. SympNet vs. PINN for learning pendulum dynamics from 40 observations. The PINN recovers physical parameters within 0.1% PyTorch
ToolScout A proposed search-then-call gateway that lets agents use hundreds of MCP tools without loading them all into context (design stage) MCP

Connect

LinkedIn

Popular repositories Loading

  1. Final-Project---gaussian-naive-bayes- Final-Project---gaussian-naive-bayes- Public

    Jupyter Notebook

  2. Wanderly-AI-Travel-Agent Wanderly-AI-Travel-Agent Public

    Wanderly - AI Travel Agent: a ReAct trip planner with an MCP tool server, pluggable LLMs (Claude, DeepSeek, any OpenAI-compatible API) and free travel data.

    JavaScript

  3. Learning-Dynamical-Systems-From-Data Learning-Dynamical-Systems-From-Data Public

    Learning pendulum dynamics from 40 observations with FNN, SympNet, and PINN in PyTorch — comparing rollout accuracy, energy conservation, and physical parameter recovery.

    Jupyter Notebook

  4. SP500-Research-Agent SP500-Research-Agent Public

    Python

  5. search-engine search-engine Public

    A2Z Search: a from-scratch Python search engine over 51K UCI webpages — SPIMI inverted index, compressed postings, BM25 ranking, and a concurrent HTTP API.

    Python

  6. AlphaLens AlphaLens Public

    Evidence-first US equity research workbench. Experimental open-source beta; no automated trading.

    TypeScript