Bio | Portfolio | In the Media | Personal Interests
This a highlight reel of the products I’ve designed and built. I made this catalog for as much you as for my own record-keeping. These projects span web, mobile, and hardware.
<aside> ℹ️ Navigation Tip: Jump around via links on the left. Images expand when clicked.
</aside>
Table of Contents:
A text-based adventure game powered by GPT-4
A Map-Maker of Phone-Based AI Agents
Designed & coded a GPT & DALL-E AI Superhero Generator web app. ****
Coded a command-line LLM-Powered AI-Code-Documenter.
An all-knowing AI chatbot that builds hack-a-thon and start-up teams based on compatible interests and skills.
Designed & coded software that enables organizers to throw 1000+ participant virtual events over a phone hotline.
Designed & coded a Google Docs alternative that uses votes to determine which edits go live. ****
Built prototypes of an electronic wearable designed to measure daily hugging behavior.
Designed & coded the prototype of a location-and-relationship-aware contacts app.
Designed & coded the winning app submission to the 2014 PCORI App Challenge, which enables researchers and patients to co-design participatory research studies.
Designed a digital health platform which enables people with Crohn’s & Colitis to get support and contribute their data towards collective medical knowledge.
Designed & coded prototypes of a mobile app that enables users to chat with those around them.
Coded front-end web implementations of design mock-ups.
Designed & coded a social network and co-housing ops software for Delta Upsilon fraternity. Won DU International’s best website award*.* ****
Contribute data for the cure.
Projects for which I don’t have portfolio pages:
TI-PROGS
Designed, coded, & distributed TI-83+ apps (utilities & games) online.
1000MEMORIES
Coded web features enabling the upload, management, and display of photos and video. ****
WORLD CLASS SHOWS
Designed & coded a marketing website and credit card payment flow.
OPENPPRN
Designed & directed an open-source project, allowing anyone to launch an online patient-powered-research-network. ****
DAPLIE
Coded a blockchain contract to manage a federated network of network-attached-storage devices. ****
AIRSPEED
Designed & coded a prototype microcontroller to measure & publish wifi speeds of AirBnBs. ****
I am currently building a text-based adventure game powered by GPT-4. I’m building it in React & Node. While it’s not yet ready for release, here’s the Development Preview Link. To start, you’ll need to log in with the authorization provider of your choice. (Note I renamed the app from TextRPG)
As a part of a two-day work trial, I built a web app that functions as a cartographer of AI voice agents. Magellan sails the high seas of the phone network to discover and map out unknown call agents. You provide a phone number, and Magellan explores the agent and returns to you a flowchart of phone menu paths.
Note that given the two-day limit, I focused on back-end functionality and then threw up a basic UI that got the job done.
.png)
Type in a phone number. Optionally provide a brief description of the business being called (for later organization) and optionally specify a maximum number of calls Magellan should make (to prevent runaway costs if the phone tree is extensive beyond expectation). Press submit. Feel free to repeat the form right away with a different number to parallelize mapping other phone agents.
.png)
A sequence of calls in progress will begin to appear, each navigating the phone tree in an incrementally different way. When each call is complete, an audio recording and transcript become available for inspection. After the first few calls, Magellan will run the transcripts through an LLM to generate a flowchart of the known phone tree paths.
.png)
The flowchart will update periodically as new calls are completed. The flowchart will display itself as complete once the LLM has deemed that all phone menu paths have been traversed or the optional max number of calls have been reached.
Tech Stack
Algorithm
I instruct the Hamming caller through performing Depth-First Search, moving left to right through the branches (using alphabetical order of possible imagined answers to each of the phone agent’s questions). Through this approach, the system caller can move with complete efficiency through the call graph, never repeating. I used an LLM call at periodic intervals (default: every 5 calls) to evaluate the transcripts and distill them into a flowchart. When that interval passes again, the LLM updates the mermaid (flowchart) diagram with the new paths exposed in the latest transcripts. If it deems all paths traversed, the system completes.
I built Hero.Rodeo, a web app that uses OpenAI’s GPT and DALL-E to generates superhero “trading cards” from simple user prompts. It generates amazing heroes from prompt text consisting of animals, fantasy characters, creatures, and objects! Try it Out!
I built it using Python, Flask, TailwindCSS, Amazon S3, and OpenAI’s GPT and DALL-E API. I designed the UX on the fly as I developed. One thing I noticed early on in user testing is how much near-instantaneous joy people got from the app. They provided an animal, eg. “squirrel”, and then seconds later “Master Nutcracker the Squirrel” appeared. When I user-tested in social settings, users would often turn to their friends to show them what they just made. This learning helped me see the value of enabling the heroes that others had created to be shared and discovered by others.
I gave each hero a unique shareable link and added navigation that allows users to flip through a gallery of “hero trading cards” in the catalog.
.png)
.png)
.png)
I worked with Tyler Neylon on coding a GPT script that automatically augments files of code with AI-generated code documentation.
It currently runs as a command-line tool which runs through your code, inserting GTP-generated python docstrings into your files and functions.
https://github.com/tylerneylon/add_docstrings
As an input argument, provide a file containing undocumented code.
Mark is the all-knowing super-connector who knows everybody present at your hack-a-thon, their skills, interests, project ideas, and need for teammates. He can suggest highly compatible team compositions, naming the project, each member and their area of contribution. You can ask him innumerable questions to get matched with the right people to work on your project idea.
Mark uses Retreival-Augmented-Generation (RAG) via a vector database and GPT4 to store the DevPost profiles of all the participants at your hack-a-thon, and query them like the all-knowing super-connector he is. I built this as a team of one, a solo hack-a-thon entrant, for the RAG Hackathon on 2/5/24 in Santa Clara, CA sponsored by LLamaIndex.
Out of the 50+ projects demo’d, this project got the only full-audience-spontanenous-ovation, earning it the unofficial audience-choice award.
My inspiration to build this came from the fact that the hack-a-thon itself was not well-set-up to allow people who arrived not yet having a team to find and join teams to build projects for the hack-a-thon. So, somewhat cheekily, I decided to make my project be the solution to the problem I had, which was not having an efficient mechanism to be match-made with other developers who wanted to work on a particular problem, and had the skills to develop it with me. Therefore, I built that exact thing, somewhat aptly, as a team of one.
