Hi,
My name is Irwin Deng.
Welcome to my |

This is a song recommendation engine that makes use of the Spotify API. Features ● Look up songs based on album, name, or artist ● Display song statistics such as charts it appeared on, streams, tempo, time signature, etc. ● Display similar songs based on those statistics ● Search for albums, and see all the songs in that album as well as similar albums ● Search for artists, and see collaborators, as well as trending news articles about them ● Look up historical top charts and display songs that were on that chart ● Leaderboard of artists who have had the most streams, as well as time on a top chart ● Account Creation/Login: Allow users to favorite songs to come back to later

For this project, I and two other students in my machine learning course studied the problem of OCR, specifically the recognition of LaTeX formulae. First, we synthesized a training dataset of over 100 thousand simulated handwritten equations. Then, we trained and tested 3 different machine learning models on that dataset, finding that a capsule neural network was best able to come up with the original LaTeX commands.

This is a distributed cloud platform similar to Google Apps. This platform has a webmail service, analogous to Gmail, as well as a storage service, analogous to Google Drive. Users connect to frontend servers with their browsers. Frontend servers don't store any state; instead, all state is stored in backend servers. Features ● Frontend servers that support persistent connections, cookie handling, and chunk transfer encoding ● Load balancer that keeps track of server status and redirects clients between servers ● Distributed key-value storage with replication and guarantees on consistency and fault tolerance ● User accounts with account creation, login, and password reset ● Webmail service that can send/receive emails internally as well as to remote users ● Storage service that allows creating, deleting, renaming, and moving files/folders ● Admin console that displays the status of servers and allows for killing/starting individual servers

This is a UNIX-like operating system simulator with a priority scheduler, FAT file system, and shell. Unlike a real operating system, this does not boot on hardware; rather, PennOS runs within a single process on a host OS. Features ● Processes that run as context threads on the host OS ● Signal handling, including SIGSTOP, SIGCONT, and SIGTERM ● Priority-queue-based scheduler with multiple process priority levels ● FAT16-based file system supporting creating, deleting, renaming, and moving files/folders ● Shell with redirection and builtins such as cat, ls, touch, chmod, and kill ● Script execution

This is a compiler that will compile code written in J, a stack-oriented language, into assembly code for LC4, an ISA created by Penn. This compiler uses the same calling convention as lcc, so J programs compiled with this can call subroutines compiled by lcc and vice versa. Implemented functionality ● Decimal and hexadecimal literals ● Arithmetic operations (+, -, *, /, %) ● Comparisons (<, <=, ==, >=, >) ● Logical operations (and, not, or) ● Stack operations (drop, dup, swap, rot, argN) ● If statements (supports nesting) ● Comments ● Functions

Java application to find the shortest path between airports, taking into account the Earth's spherical geometry.

This is an application I wrote for my high school marching band to automate the creation of dot books, which we used to learn formations on the field.

This is the second iteration of my portfolio website, written in TypeScript + Angular.