Skip to content
Gavin Lam

Projects in depth.

What each project is for, what it does, and how it is built.

Concept sketch with sample data, not a screenshot.

Xur

Voice assistant with a 3D holographic map table, 2026

The problem

Maps are usually flat. Xur is meant to feel like the holographic map table from a movie: you ask for a place and it rises up in front of you.

What it does

An AI assistant that helps users navigate, explore and plot routes on maps. Built with React and Tailwind, it uses LLM APIs to give real-time guidance.

Key features

  • Voice-interactive AI map guide
  • Ask it to show a place and the place appears in 3D
  • 3D route mapping

Built with

  • React
  • JavaScript
  • Tailwind
  • Gemini API
  • Mapbox API

Concept sketch with sample data, not a screenshot.

Wrenchlog

In development

Maintenance tracker for my own car

The problem

I run two sets of tires, all-season and winter, and need to know when to swap them and what winter maintenance is due. I built this for my own car.

What it does

A vehicle maintenance tracker with winter maintenance and tire tracking built in. A weather integration watches the forecast and raises an alert when it is time to swap tire sets.

Key features

  • Maintenance tracking, including winter maintenance
  • Tire tracking for two sets: all-season and winter
  • Forecast-driven tire swap alerts through a weather integration

Technical decisions

  • Python and FastAPI backend with SQLite for storage, and a React frontend
  • Backend tested with pytest
  • Authentication is planned for a later version

Built with

  • Python
  • FastAPI
  • SQLite
  • React
  • pytest

Concept sketch with sample data, not a screenshot.

AVA (HealthChat)

Voice chatbot for mental and physical health, 2025

What it does

AI-powered voice chatbot that helps users talk through mental and surface-level physical health concerns. It started as a team project where the "chatbot" just matched input against a fixed file of questions and answers. I rebuilt it on my own around real AI-driven conversation logic, and reworked nearly everything else along the way. It is containerized with Docker and served by a Python web app in the browser.

Key features

  • Voice-interactive AI health assistant
  • Covers mental and surface-level physical health concerns
  • Runs in the browser, served by a Python web app
  • Containerized with Docker

Technical decisions

  • Replaced the original scripted Q&A lookup with AI-driven conversation logic
  • Built the conversation pipeline with LangChain, calling Google Gemini as the model

Built with

  • Python
  • LangChain
  • Gemini
  • JavaScript
  • HTML/CSS
  • Docker

Concept sketch with sample data, not a screenshot.

FitTrack

In development

Nutrition and workout tracking in one app

The problem

I wanted MyFitnessPal Premium-style nutrition tracking and a workout tracker in a single app.

What it does

A web app that combines nutrition tracking in the style of MyFitnessPal Premium with a workout tracker and a full program generator.

Key features

  • Food data from USDA and Open Food Facts
  • Workout tracker alongside the food log
  • Full training program generator

Technical decisions

  • Built on Django REST Framework and React, a stack I chose in order to learn it
  • Single user for now
  • Web app only: no wearable syncing or live GPS

Built with

  • Django REST Framework
  • React
  • USDA food data
  • Open Food Facts

Scrum, CI/CD, containerized

ToDo: Full-Stack Task Manager

Full-stack task management platform, Mar 2025 – May 2025

What it does

Full-stack task management platform built with React and Node.js following Scrum methodology. Features RESTful APIs for task creation, prioritization, and user persistence, containerized with Docker.

Key features

  • RESTful APIs for task creation, prioritization, and user persistence
  • Built following Scrum methodology
  • Containerized with Docker

Built with

  • React
  • Node.js
  • TypeScript
  • PostgreSQL
  • Docker
  • Jira

~90% classification accuracy

PUBG Player Performance Prediction

ML pipeline predicting top 10% match placement, Dec 2025

What it does

ML pipeline processing 6.3M+ PUBG player records to predict top 10% match placement. Benchmarked 10+ models including LightGBM, Random Forest, ANNs, SVM, and CART with hyperparameter tuning.

Key features

  • Processes 6.3M+ PUBG player records
  • Predicts top 10% match placement
  • Benchmarks 10+ models: LightGBM, Random Forest, ANNs, SVM, CART and more
  • Hyperparameter tuning

Built with

  • Python
  • PyTorch
  • Pandas
  • NumPy
  • Google Colab

Novel 'such that' SQL clause

Extended SQL Query Engine

Multi-pass query engine with a new 'such that' clause, May 2025

What it does

Built a multi-pass query engine in Python to simulate SQL-style grouped aggregation using dynamic predicate evaluation and in-memory data structures. Designed a novel 'such that' clause extending SQL's expressiveness for advanced conditional logic across grouping variables.

Key features

  • Multi-pass evaluation of SQL-style grouped aggregation
  • Dynamic predicate evaluation over in-memory data structures
  • A 'such that' clause for conditional logic across grouping variables

Built with

  • Python
  • SQL
  • PGAdmin

80% faster environment setup

Enterprise CI/CD Pipeline

Perforce, Swarm and Jenkins on cloud Linux, Sep 2025 – Dec 2025

What it does

Deployed and configured 3 integrated DevOps services (Perforce, Swarm, Jenkins) in a cloud Linux environment. Reduced manual build process from ~5 minutes to under 30 seconds per commit.

Key features

  • Three integrated services: Perforce, Swarm and Jenkins
  • Runs in a cloud Linux environment
  • Manual build cut from ~5 minutes to under 30 seconds per commit

Built with

  • Jenkins
  • Docker
  • Perforce
  • Linux
  • DigitalOcean

Let's build something together.

Actively seeking full-time software engineering roles: full-stack, backend, or ML/AI. I reply within 24 hours.