About
Hi there! I am Daniel. I work as a principal research engineer, working on problems related to augmented-intelligence and building better tools for scientists/researchers.
I am primarily interested in Visualization, Machine Learning, and HCI-related topics. This page lists some of my side projects. I am always up for collaboration on interesting ideas and learning new things. Drop me a line at mwdchang@gmail.com.
Recent projects and works in progress
Boids simulation
Agent based modelling with boids simulation. Written in Javascript and webGL
Point cloud volume
Point cloud enclosed within parametric volumes. Made with ThreeJS and Tweakpane. This came from a discussion about using a custom shader pipeline backed by textures vs vanilla memory buffers for points transforms. This is the memory buffer version, it works pretty well (pending on enclosed geometries) to about 1 million particles on a 2013 MacbookPro.Machine-Learning and AI
Hackathon 2023
We made Azul, a Slackbot capable of summarizing and synthesizing conversations, as well as providing additional context. Imaging the case where you are back from a short vacation, or you are not familar with the topics of dicussion, Azul can help!! Azul is built on top of GPT3.5 LLM
Tesseract OCR
A handy web-tool for extracting text from clipboard. For those times when you have to deal with other people pasting images into Slack instead of text ... :)Better search
Experiments with better ways to search through a collection of text, with intentions to extract out co-occuring narratives with NLP/ML techniques.
GAN
A simple Generative Adversarial Network for generating hand-written digits. Written in vanilla tensorflow-js without any additional 3rd party libs. Great for understanding how GAN works.
LDA
Using LDA for topic modelling against famous quotes
Wine classifier
Experiment with simple wine classification using TensorflowJS
Q-Learning
Navigational solver with Q-Learning technique
Text generation
Text generation with LSTM and Maximum-Likelihood models
Louvain detection
Louvain community clustering with force-directed layout
Batter classifier
Tensorflow classification for MLB batters based on career statistics. Divide batters into 4 different hall-of-fame categories and see if we can build a classifier based on simple stats.
DBSCAN
DBSCAN clustering algorithm and visualization (neighbourhood growth).
K-means
K-means clustering algorithm and iteration (center seeker) visualization
SOM
Experiment with Self-organizing map. Here we encode MLB batters's status to detect similar players.Data Visualization and Graphics
Computer graphics and data-viz related things. I have a particular interst with procedurally generated images and image synthesis.
TSP-Fourier
Image synthesis with TSP and Fourier transforms. Convolve an image and reconstruct it as a graph, then reformulate it as Fourier transforms.
Song Visualization
Visualizations and analyses of team's spotify playlist: track attributes such as length, energey, popularity, plus analysis of lyrical similarity
Song Tiles
Tile mosaic from album covers. Divide an image into grids, and for each cell calculate the best matching tile from a pool of images.
Polaroids
Polaroid shader effects
The Other Side
Finding and visualizing antipodes
Attractors
Strange attractors, chaos systems, and other procedurally generated things
Uncharted Hackathon 2018
Pollution release correlations
Arc diagram
Using arc diagram in place of venn diagrams
Fisheye shader
WebGL Fisheye effect
Perlin noise
Generating 2D textures with Perlin noise superimposed onto random functions
WebGL Shaders
Testing Three.js shaders via procedurally generated plasma texturesTinkerings
Miscellaneous experiments.
Workbench
A research synthesis tool for organizing, annotating, and presenting research artifacts. Inspired by a real-world "workbench" metaphor with 2D-physics
Discord bot
Analyzer bot for the online game the-reincarnation. The bot interact with users, and performs battle simulation, match analysis, and suggest optimal unit pairings. 
Image Utilities
A collection of image manipulation functions for working with ImageData object. Colour filters, convolution kernels and other interesting effects
Watts-Strogatz graph generation
Straying into graph theory. Experimention with generating and visualizing graphs with "small-world" properties
Baseball Hackathon 2018
The Opinionator. Third place winner in the hackathon!!
Fountain code - LT
An interactive visualization of Fountain code encoder/decoder using Luby transformations.
Stckr
A flexible, API driven stacking library in D3
My thesis project
VehicleVis: combinging text visualiation with 3D graphics to investigate complaint reports
Clock
Abstract and artistic clocks written in D3. This utilizes the notion of arcs and positive/negative spaces to represent time segments
Grid reduction experiment
Reduce grid noise by only showing localized context around the mouse cursorUnfinished
These projects are, for one reason or another, incomplete or abandoned. Nevertheless, there are still some interesting/good ideas, and code that can be reused/recycled for other shenanigans.
Publications
Research publication I was involved in.
- ACM-CHI 204 (Workshop): Towards Accessible, Flexible and Collaborative Scientific Modeling with Terarium
Proulx, P., J. Waese, MW. Chang, et al, - IEEE-VIS 2021: A Mixed-Initiative Visual Analytics Approach for Qualitative Causal Modeling
Fahd Husain; Pascale Proulx; Meng-Wei Chang; Rosa Romero-Gómez; Holland Vasquez - MobileCHI 2014: Interaction for Reading Comprehension on Mobile Devices
Rafael Veras, E Paluka, MW Chang, V Tsang, F Shein, C Collins - Pacific-Vis 2013: Exploring entities in text with descriptive non-photorealistic rendering
Meng-Wei Chang, Christopher Collins
This page is made haphazardly with Vue3, Boostrap, and D3. Stirred together until the number of JS errors approached 0 ... ¯\_(ツ)_/¯