Hirschibar: Code
Welcome to my Github Portfolio Overview!
On this page you can find out what projects I’ve been up to. Feel free to follow me on github or to reach out if you’d like to collaborate! You can also find me on linkedin and twitter.
Here is my resume (Updated Sept 2023) if you want to see it.
And now, on to the portfolio…
Company Update
Amira Invest
Update: Sadly, this company has been shutdown as of summer ‘22. My co-founder and I decided to split up, and life (aka the arrival of my second child) required I step away and get a “real job”. It was a tough decision, and while it marks the end of Amira Invest, it certainly is not the end. Thanks to everyone for your support, I’m sorry it had to come to a close.
Parts of the code have been made public so you can see what we did. Check it out
I’m building a marketplace platform that provides the structure and the incentives needed for you to trust the investment opinions you find online.
The project includes over 300 PRs, and 1200 commits with 5 contributors. Follow us here so you can join as we open source components of our app. Sadly, this code is unable to be shared publically yet. But stay tuned and follow here! There may be changes soon! 😉
I’m currently working on a recommendation engine which builds a unique model for each user to recommend content of most worth and relevance to each user’s feed. This article from Meta’s engineering blog will give you a sense of what I’m building.
Join our waitlist at https://amirainvest.com if you’re interested in managing your own investments with more confidence!
Current Projects
Option-bot
I’m building a reinforcement learning agent to learn to manage a portfolio of short option contracts. I.e. to learn when to hold, close out, or roll my short positions deterministically. Using Polygon data, and building on the Pyrh (unofficial robinhood api) project to build this bot. Open to collaboration! Check out the code here
Basically, trying to intelligently capture theta decay without crumbling to the emotions of the moment like I do personally. The bot will learn a policy based on probabilities and gain the experience of a skilled trader.
The goal is to generate consistent income, targetting $100/day of short option premiums. Interested? Reach out!
Boardwalk
Are you sick of walled gardens like I am? In this day of devices and data, we are told that our data is ours, yet it often remains locked in some company’s app where you can’t do anything with it. Lame!
This project is building the Plaid of wearable health data, enabling developers to build apps that access data from platform. Gone are the days where, if your friends buys a different watch, you can no longer enjoy step challenges or other social fitness features.
Breaking down walled gardens one brick at a time, supporting the Open Data Movement!
Past Projects
Academy Summarized
(Grad School Project) I co-authored a paper with colleagues at GaTech where we analyzed the effects of fine-tuning an encoder-decoder transformer on State-of-the-art pretrained models using Arxiv academic paper data. We found that by extracting significant sentances from the papers, we could auto-generate summaries/abstacts better than with the non-fine-tuned version.
Below is our abstract
Music Box: a better Spotify playlist generator
(Grad School Project) I co-authored a poster with a team at GaTech that built an app to generate better Spotify playlists using clustering algorithms. Read about or contribute to the project here, or try out the app in your browser here
Here is a pic of the UI from the app. Click here for the full poster:
Scratch ML: Machine Learning Algorithms written from scratch
Here is where I’ve been collecting tools (algorithms) I’ve built completely from scratch as an exercise to make sure I understand them.
Initial efforts start with only using numpy and python. Once the foundations are established, I will later make use of PyTorch or other tools to build out other tools. This is a new work in progress (idea came over a year ago, which I’m just barely picking back up, we will see if I actually do anything with it this time).
Give a follow, or DM me if you’d like to contribute. It’d be sweet to get a out-of-the-box, easy to use set of Machine Learning tools prepped into a PyPi package that learners could experiment with.