michaelwan.dev

What It Is

michaelwan.dev is the personal homepage of Michael Wan. It is positioned more like a résumé and portfolio for a data scientist / AI specialist than a register-and-use AI application. The site highlights his experience at companies such as Mastercard and OOCL, with a focus on deep learning, LLMs, RAG, enterprise automation, fraud detection, and large-scale operations research optimization.

Core Capabilities

Based on the page content, his strengths are centered on enterprise AI implementation. At OOCL, he led an LLM + RAG + LangChain-based customer service automation project that reduced email response time from 5–10 minutes to a few seconds. He also built a GenAI-powered API migration tool covering 20k+ customers, with a claimed 82% accuracy rate and 8,000+ person-days saved. Other areas mentioned include LightGBM/PyTorch forecasting, CPLEX optimization, reinforcement learning, supervised learning, and genetic algorithms. His Mastercard experience includes fraud detection platforms and Databricks ML pipeline optimization, improving training speed by 3x.

Pricing and Integrations

The site does not provide product pricing, a free trial, an account system, API documentation, SDKs, or external integration entry points. As such, it should not be considered a standard AI SaaS tool. Although the page mentions technologies such as LangChain, Databricks ML, J2EE/Oracle, MEAN, and EDI/API migration, these refer to past project tech stacks rather than callable services currently offered by the website.

Pros and Cons

The main advantage is that the project examples are concrete and include business metrics, covering high-value scenarios such as financial risk control, shipping optimization, and customer service automation. This demonstrates strong experience in enterprise AI and optimization modeling. The drawbacks are also clear: there is no hands-on product, demo, privacy policy, service support, Chinese-language support, or commercial purchase path. The results are primarily described as part of an individual career history, making them difficult for external users to verify or reuse.

Who It’s For and Access from China

This site is better suited for recruiters, enterprise AI project leads, fellow data scientists, or readers interested in real-world LLM/RAG implementation cases. If users need ready-to-use alternatives, they can consider LangChain, LlamaIndex, Dify, or Flowise for RAG/Agent development; ChatGPT, Claude, or Gemini for general AI; and CPLEX, Gurobi, or Databricks ML for optimization modeling. The page does not state access conditions from mainland China, and network availability and payment information are unknown.