Data Science DC July 2024: 'Using LLM Embeddings for a Dating App' and 'Python As an R User'

From Alex Long: “Using LLM Embeddings to Build a Recommender System for a Dating App Based on the News”

From Marck Vaisman: “The Things About Python That Drive Me Mad (As an R User)”

Descriptions

Using LLM Embeddings to Build a Recommender System for a Dating App Based on the News

News consumption can reveal shared interests and perspectives between individuals. This talk covers how InPress leverages large language models and clustering techniques to match users romantically or platonically based on their news engagement patterns. We’ll discuss generating rich user representations from news data using fine-tuned LLMs, clustering these representations to identify groups with similar news interests, and efficiently matching users within these clusters for friendship or dating. Insights from data we gained in a recent beta test and used to evaluate our approach will also be presented.

The Things About Python That Drive Me Mad (As an R User)

In my life as a data person, I started with and mastered R (I even ran R on Hadoop back in 2010-2012 timeframe...). However, over the years (especially at work), I have been using Python more and more and really trying to understand the mental model and best programming practices. This talk is **NOT** about making an argument for which one is better, or trash talking about either (well, maybe a little for fun's sake). It's about the gripes I've had when trying to do tasks in Python that are (in my opinion) less intuitive. I do believe that today's data people should know enough about both and have a more extensive toolkit.

[brianholland]

I'd like to volunteer to ensure that the sound/audio on these recordings matches the high quality of these presentations. I live in DC, just graduated from Georgetown in Data Science for Public Policy, and am currently looking for data work (and have applied at Excella). Who would be the best person to get in touch with about that?