Jan Frederik Mohr

I’m a graduate of ETH Zurich, where I completed my bachelor’s and master’s degree in Electrical Engineering and Information Technology.

I’m highly curious and deeply fascinated by machine learning and data science applied to the financial markets.

Below, a collection of some of my personal projects is displayed, ranging from an interactive dividend dashboard, to an LLM-powered trading bot, to cloning actual doctors in their appearance and knowledge with an LLM and human-like avatars.
Enjoy!

Connect with me via email or LinkedIn:

jan.frederik.mohr@outlook.com
https://www.linkedin.com/in/jfmohr/

PatientEd

Context

  • For high-volume operations like lasik eye surgery, doctors have to explain the same process to each patient and therefore need to repeat themselves a lot, which costs them a lot of valuable time that could have been used for more in-depth conversations with specific patients.

Idea

  • Creating a digital human-like avatar of the doctor and using an LLM that is grounded on a specific type of operation, like lasik eye surgery, the digital twin of the doctor can take over the repetitive tasks, saving him a substantial amount of time.

  • A few large clinics calculated that this solution could save them the time of one full-time doctor.

Solution

  • The video below shows a prototype version, which I programmed. We are in the role of Ms. Miller, who is being educated on lasik surgery.

  • The avatar answers with LLM-generated responses to our questions, specific to the type of operation.

Interactive Dividend Dashboard

A small project which displays key dividend metrics of companies in the major stock indices.
It was programmed in Python and the Streamlit library was used to create the front-end.

LLM-Powered Trading Bot

Context

  • There are Youtube channels with 500’000 followers or more, who sometimes post videos titled “The top 10 cryptocurrencies to invest in right now!”.

  • Subsequently the price of these cryptocurrencies rises up to 10% in a single day since thousands of people start buying them, directly after the video is released.

Idea

  • If we would directly know which cryptocurrencies are positively talked about, we could be the first one to buy them on the exchange and subsequently profit from the price increase.

Solution

  • I built a Flask web server application that subscribes to Youtube channels using POST requests to get immediately notified when a new video is uploaded.

  • Using the youtube-transcript-api library in Python, we can obtain the complete text that is spoken throughout the entire video as a string.

  • An LLM subsequently analyzes the entire string and outputs us a list with each cryptocurrency that was mentioned and a number of 1-10 specifying how positively it was mentioned.

  • Directly afterwards, the Python script executes buy orders on the Binance exchange for each cryptocurrency rated 8 or higher.