A project walkthrough to propose the best incubators for start-up founders, using Python, Pinecone, FastAPI, Pydantic, and Docker
Harness, a startup dedicated to assisting founders in their entrepreneurial journey, approached me to develop a tool that aids their community in finding the most suitable incubators: the Matching Tool.
In this article, we walk through the different stages of this project, from the solution design to the delivery.
The company and its cofounders would like to create a tool that enables their community of start-up founders to find the best incubators & accelerators around the globe.
To do so, they manually collected data from incubator websites, including details such as location, various requirements, funding opportunities, and more. Additionally, they leveraged an engaged community of founders.
With the data from incubators and their community, they needed to find a way to retrieve the top-k incubators based on start-up information.
Challenge accepted.
Overview
At first glance, the project looked like a Recommender system like Netflix or Amazon used to suggest the best series or products to their users. From user behavior such as clicks, reviews, or upvotes, a company can anticipate and recommend the most suitable product.
Yet, in this particular scenario, we lacked any prior data on a founder’s preferences. Thus, building a Recommender System was unfeasible in this case.
An alternative approach could have involved embedding incubator and startup data into a vector space for a similarity search. Put simply, this method entails measuring the distance between vectors to identify the closest incubators in proximity to a given startup.
But this approach had many defaults in this case.
Incubators have what I call hard criteria, factors that could result in immediate…
