AI, Data Science & Product

From Problem to Product.

I’m Lukas. I build apps and tools that help people plan their day, compare costs, and work with AI. I shape the product, direct AI-assisted development, and check the result in use.

Independent Projects by Xu Zhou (Lukas)Mobile · Web · Desktop · Developer Tools

A Few Places to Start

Selected Work

Three projects that show how I approach product design, data modeling, and AI-assisted development.

Explore All 11 Projects

Consumer Apps · AI Tools · Knowledge & Productivity

How I Build

I use AI agents for implementation and review. I’m responsible for the product direction, design choices, and release decisions.

01

Define the Problem

Start with the person using the product and the decision they need to make. Identify the assumptions, data limits, and smallest useful workflow.

02

Build for Change

Keep product rules separate from the interface and services. Record trade-offs, protect saved work, and reuse code when the same need appears across apps.

03

Test the Result

Check the result through tests, independent review, and the installed or deployed app. In Car Shopping, review caught an unsafe URL import; I removed it and kept a regression test.

What I Bring

I connect product decisions with data analysis and practical software delivery.

AI Workflows

  • Scope tasks and direct specialized agents
  • Define interfaces and check generated work against sources
  • Turn verified lessons into reusable workflows

Data Science

  • Python, pandas, scikit-learn, and XGBoost
  • Predictive modeling, clustering, and topic modeling
  • Explain model assumptions and uncertainty

Product Judgment

  • Define the problem and prioritize a useful first workflow
  • Turn complex information into comparisons and next steps
  • Weigh usability, privacy, cost, and release risk

End-to-End Delivery

  • Mobile, web, and local desktop products
  • Separate product rules from storage and the interface
  • Check releases and make remaining limitations clear

The case studies show how I apply these skills, including the checks that caught problems and the limits of each result.

Earlier Work

Earlier work in machine learning, data analysis, and AI applications.

AI Agents and LLM Apps

3 Projects

Experiments in machine learning and early AI applications.

  • AI Agent Experiments

    ML and reinforcement-learning experiments (a DQN Lunar Lander, a config-driven ML pipeline, an agent chat app) built with an AI coding agent.

  • AI Builder Mini-Projects

    Five Python mini-projects from an AI-builder course: crypto price plots, a CVPR paper scraper, GUI automation, image analysis.

  • Search-Grounded Chat Apps

    Early-2023 Streamlit experiment: ChatGPT answers grounded in Google search results, plus a scheduled news-summary emailer.

Finance Experiments

1 Project

Data Science

4 Projects

Notebook projects in predictive modeling, clustering, and text analysis.

Continue the Conversation

Let’s Work Together

I’m interested in opportunities across AI, data science, and product development. If the work here is relevant to your team, I’d be glad to discuss the problems you’re solving.