Projects

SentGen is a computational linguistics program that is meant to stochastically generate gramatically correct English sentences. The generator works by taking in a specified Engilsh grammar in the form of phrase structure rules, and building a tree structure whose root is a 'Sentence' node and whose leaves are English words (terminal symbols). Optional rules are followed randomly to produce some variations in sentence structure.

The English word data was generated by unifying the Google 10k most common English word dataset with a 370k Part of Speech tagged English dataset. This program is a computational implementation of the Phrase Structure theory presented in UCLA Prof. Bruce Hayes' Introductory Linguistics. The grammar used in the program is also directly from this text. Work still needs to be done to add proper articles, prepositions, and inflection. Semantics are completely ignored, so the sentences will likely not make sense.

My first polished commercial product, Starbox is a dynamic light display that continuously generates random star clusters of different sizes, colors, and brightness. Rare shooting stars are also generated and displayed. Device enclosure was modeled in Fusion 360. Star generation and HUB75 light display written in Python. Available now at my Etsy shop linked below! As of February 2023, we have sold 5 Starbox units!

Using the University of Pennsylvania SIG Lab's Vicon Motion Capture System, I recorded three separate MoCap scenes. Using the captured data and Autodesk MotionBuilder, I applied the data onto a rigged character model, and then blended the transition between the three scenes so that the result is a realistically continuous animation.

Back during my sophomore year of high school, I attended the University of Pennsylvania's hackathon, PennAppsXV, and worked on a device that swiftly opened a rotary Masterlock based on digital keypad input. I 3D printed a piece that latched onto both the shaft of a stepper motor and the dial of the Masterlock. With an Arduino, I was able to translate digital keyboard input of the lock's combination and convert that into corresponding rotations that opened the lock with great speed. In typical hackathon fashion, the enclosure for the device was intricately constructed out of cardboard and electrical tape.

Released to the App Store under my ed-tech company Tryber Inc., my co-founder and I developed EasyA in order to provide free study tools for students who struggled with maintaining structured study habits. I developed the customer-facing app and internal backend, while my co-founder wrote the course content and exercises. Once the user marked that they had learned a topic in one of our three supported courses (Calculus I, Intro to Biology, Intro to Chemistry), EasyA would notify the user in spaced intervals to review a summary of a topic they had learned, as well as take a short quiz to test their comprehension.

I wired up an accelerometer, an individually addressable LED strip, and a 6V battery pack. Then, I modeled a closed case for this arrangement in Blender, 3D printed it, and screwed it to the bottom of my board, where its other electronics lay. The lights are fully programmable, and can react to the motion data captured by the accelerometer.

Given an input of keyframed joint angles for various animations, I implemented the Forward Kinematics and their interpolation for a rigged character model. Two Inverse Kinematics techniques were also implemented, Limb-Based IK and Cyclic Coordinate Descent (CCD) IK to allow joint angle computation given a target position of a joint. This project was assigned in my Computer Animation course (CIS5620). Source code is privated due to course integrity restrictions. Feel free to email hi@aykae.com for access.

Implementing the node-based programming of SparkAR, combined with the various Penn State College frames I designed in Photoshop, I was able to create a roulette type animation that concluded with a randomly selected college for the filter user. The first filter published to Onward State's Instagram, it garnered roughly 100,000 impressions via 50,000 videos created.

Intended to show the collective staff prediction of Penn State's College Football Playoff chances based on the variable outcome of their next three games, this Javascript app was the first of its kind to be published on the Penn State blog, Onward State. Developed using JavaScript Canvas and Adobe Photoshop, this app allowed Penn Staters to simulate all possible outcomes of their football team for the tailend of the 2019 season.

SentGen is a computational linguistics program that is meant to stochastically generate gramatically correct English sentences. The generator works by taking in a specified Engilsh grammar in the form of phrase structure rules, and building a tree structure whose root is a 'Sentence' node and whose leaves are English words (terminal symbols). Optional rules are followed randomly to produce some variations in sentence structure.

The English word data was generated by unifying the Google 10k most common English word dataset with a 370k Part of Speech tagged English dataset. This program is a computational implementation of the Phrase Structure theory presented in UCLA Prof. Bruce Hayes' Introductory Linguistics. The grammar used in the program is also directly from this text. Work still needs to be done to add proper articles, prepositions, and inflection. Semantics are completely ignored, so the sentences will likely not make sense.

Released to the App Store under my ed-tech company Tryber Inc., my co-founder and I developed EasyA in order to provide free study tools for students who struggled with maintaining structured study habits. I developed the customer-facing app and internal backend, while my co-founder wrote the course content and exercises. Once the user marked that they had learned a topic in one of our three supported courses (Calculus I, Intro to Biology, Intro to Chemistry), EasyA would notify the user in spaced intervals to review a summary of a topic they had learned, as well as take a short quiz to test their comprehension.

My first polished commercial product, Starbox is a dynamic light display that continuously generates random star clusters of different sizes, colors, and brightness. Rare shooting stars are also generated and displayed. Device enclosure was modeled in Fusion 360. Star generation and HUB75 light display written in Python. Available now at my Etsy shop linked below! As of February 2023, we have sold 5 Starbox units!

I wired up an accelerometer, an individually addressable LED strip, and a 6V battery pack. Then, I modeled a closed case for this arrangement in Blender, 3D printed it, and screwed it to the bottom of my board, where its other electronics lay. The lights are fully programmable, and can react to the motion data captured by the accelerometer.

Using the University of Pennsylvania SIG Lab's Vicon Motion Capture System, I recorded three separate MoCap scenes. Using the captured data and Autodesk MotionBuilder, I applied the data onto a rigged character model, and then blended the transition between the three scenes so that the result is a realistically continuous animation.

Given an input of keyframed joint angles for various animations, I implemented the Forward Kinematics and their interpolation for a rigged character model. Two Inverse Kinematics techniques were also implemented, Limb-Based IK and Cyclic Coordinate Descent (CCD) IK to allow joint angle computation given a target position of a joint. This project was assigned in my Computer Animation course (CIS5620). Source code is privated due to course integrity restrictions. Feel free to email hi@aykae.com for access.

Implementing the node-based programming of SparkAR, combined with the various Penn State College frames I designed in Photoshop, I was able to create a roulette type animation that concluded with a randomly selected college for the filter user. The first filter published to Onward State's Instagram, it garnered roughly 100,000 impressions via 50,000 videos created.

Back during my sophomore year of high school, I attended the University of Pennsylvania's hackathon, PennAppsXV, and worked on a device that swiftly opened a rotary Masterlock based on digital keypad input. I 3D printed a piece that latched onto both the shaft of a stepper motor and the dial of the Masterlock. With an Arduino, I was able to translate digital keyboard input of the lock's combination and convert that into corresponding rotations that opened the lock with great speed. In typical hackathon fashion, the enclosure for the device was intricately constructed out of cardboard and electrical tape.

Intended to show the collective staff prediction of Penn State's College Football Playoff chances based on the variable outcome of their next three games, this Javascript app was the first of its kind to be published on the Penn State blog, Onward State. Developed using JavaScript Canvas and Adobe Photoshop, this app allowed Penn Staters to simulate all possible outcomes of their football team for the tailend of the 2019 season.