Practical Hacking at the Y-Combinator Hackathon
On November 22nd and 23rd, 2019, Princeton seniors Kevin Hou, Lauren Johnston, and Theodor Marcu represented the Princeton Entrepreneurship Club, a recognized student group, through an ODUS and SEAS sponsored trip to the Y-Combinator Hackathon in Mountain View, California. Lauren Johnston chronicled their experience:
The unique combination of innovative spirit and immense time pressure creates a special kind of magic at hackathons. In the span of 24 to 36 hours, participants are challenged to build something that no one has ever seen before--from augmented reality applications for doctors to cryptocurrency-based savings accounts. The Y-Combinator (YC) Hackathon is no exception to this spirit of innovation: in its first iteration in 2014, 17-year old Tanay Tandon wowed attendees with a demonstration of how a five dollar lens attached to a smartphone camera could be used as a microscope and, in conjunction with machine learning techniques, automatically detect the presence of malaria on a drop of blood. Meanwhile, Tandon’s project was not only a brilliant innovation but also the beginning of a stellar business idea. Five years later, Tandon is running a startup called Athelas that sells FDA-approved devices that can detect immune cell counts from blood samples in minutes .
Drawn in by YC’s pragmatic wisdom, our team decided to apply to the Hackathon. Upon receiving our acceptances, we immediately began thinking about potential projects. We were determined to find a project that had a strong social mission but could also turn a sizeable profit. While searching for an idea, we asked ourselves “what problems do people struggle with on a regular basis?”. Ultimately, we realized that we all had both observed and experienced the struggle of getting paid fairly for content creation--from selling photo prints online, to finding advertisers for local print newspapers, to monetizing independent content on Youtube. Given this shared experience, we decided to build an application that would help content creators earn sustainable incomes.
One area of content creation that we became particularly interested in is the podcast industry. We saw a lot of potential in the recent growth in podcast listenership; venture capital fund Andreesen-Horowitz recently reported that 65% of the active listeners started listening in the past 3 years . Furthermore, according to the same report, more than a fifth of Americans listen to podcasts on a weekly basis. Due to these promising statistics, we wanted to create a podcast platform that both emphasized fair compensation for content and had a unique way to attract listeners away from larger platforms such as Spotify and Apple Podcasts.
Building Ask Ava required us to develop several components, a question and answer (Q&A) model trained on podcasts, a podcast transcription pipeline to train the model, and a custom Amazon Alexa application that would prompt users to ask questions and communicate with a server to retrieve answers from our trained model. In order to create the Q&A model, we used a natural language processing model developed by Google in 2018 and trained it on a set of podcast transcripts. We developed this set of podcast transcripts by scraping the web for podcasts and then transcribing them via Amazon’s automatic audio-to-text transcription service. The sketch below shows the interactions of the major components of Ask Ava.
After 24 long hours of hard work, our team was thrilled to finally have a working application. We prepared a short pitch and video to present to two YC Partners. However, when we walked into our pitch meeting, it was clear that they were interested in hearing about our project in a more raw, un-filtered way. The partners asked for a live demo and a two-sentence description from each of us about how people would use our application. They were curious about how we arrived at our project idea and how we had verified that there was a market for it. Slightly surprised by this turn of events, we dropped our original pitch and complied.
The challenge of pitching our project’s business value ultimately became just as important as the process of building. After being forced to think on the fly and put our market research to use, we came away with a lesson that no other hackathon could have taught us: your project only matters as much as the world needs it. No amount of ingenious invention and incorporation of cutting-edge technology can negate the need to ascertain that there is a genuine demand for it.
 Y-Combinator Blog. https://blog.ycombinator.com/athelas-our-road-from-hack-to-product/
 Andreesen Horowitz. https://a16z.com/2019/05/23/podcast-ecosystem-investing-2019/