Skip to content
William Lei

Data, AI and web apps, built and measured.

I'm William Lei, an independent engineer in Hong Kong. I take a product from data pipeline to deployed web app with AI built in, and report the result in numbers.

Book a 20-min intro callEmail me

Python, TypeScript, Next.js, MongoDB, RAG and LLMs, AWS and Azure.

FinFAST.ai, before and afterMeasured in production

Typical query time

Before
~10 s
After
~1.5 s

6.7× faster on typical queries

Daily LLM token cost

100% to 20%

RAG retrieval accuracy

baseline to +70%

What I can build for you

Four kinds of project, each backed by work I've already shipped.

  • AI and LLM features

    Chat over your own data, retrieval (RAG) that finds the right answer, and automated multi-step agents, built to stay accurate and affordable once real users arrive.

    • A chat assistant that answers from your documents or database
    • Fixing a RAG system that returns the wrong sources
    • Cutting the monthly bill of an existing LLM feature

    Raised retrieval accuracy 70% and cut daily token cost 80% on FinFAST.ai. See the case study

  • Full-stack web apps

    Next.js and React products with the backend, login and cloud setup they need, from a first version to a fast, maintainable app your team can keep shipping.

    • A customer portal or internal tool with Google sign-in
    • Speeding up a slow dashboard or API
    • Deployment and CI/CD on AWS or Azure

    Took typical data retrieval from about 10 seconds to about 1.5 seconds. See the case study

  • Data pipelines and analytics

    Automated collection, cleaning and validation, with the dashboards and models on top, so the numbers your team looks at every morning are fresh and correct.

    • Scheduled pipelines that pull from APIs, files or websites
    • Dashboards that replace a weekly spreadsheet
    • Forecasting and statistical models you can explain

    Building scrapers for about 80 UK university course catalogues for a freelance client. See the case study

  • Mobile app launch support

    Help getting an app through Google Play and App Store review: privacy policies, data-safety answers, deletion flows and release setup.

    • Privacy policy, terms and data-deletion pages
    • Store data-safety and privacy questionnaires

Selected work

All case studies

How a project runs

  1. 01

    Intro call

    20 minutes on what you're building, what's slow or broken, and whether I'm the right person.

  2. 02

    Scope

    A short written plan: what gets built first, how we'll know it works, and what it needs from you.

  3. 03

    Build

    Small, working steps. You see real software early and often, not a big reveal at the end.

  4. 04

    Hand over

    Readable, tested, documented code your team can run and extend without me.

Have something similar in mind?

Book a 20-minute intro call, or write to william@wlei.org.

Book a 20-min intro call