Inspiration
We came into the hackathon wanting to build something that could help students and other young adults like us, we initially settled on a job opening platform to help applicants find specific opportunities tailored towards them. During this process however, we found that we could learn a lot about a company from the information the job board APIs provide and expanded our project to aid investors by providing them intelligence on the health and direction of a company based off their job listings.
What it does
Signal Foundry scrapes the job board data from two API sources, Greenhouse and Lever, it then takes the data to label and sort the specific data that is essential for our future calculations, we then store that data in our SQL database and then compute that data internally.
For investors, once our computation is complete we then send growth trends, role composition, specific signals, and other data to Nemotron then have it analyze and output a well formatted response on our webpage.
For applicants, we take the job listing data and display it to the applicant, matching the applicants inputted skills to jobs with expected future growth.
How we built it
We built SignalFoundry using a Java backend, SQLite databse, and a React/Typescript frontend, organized under our five-stage pipeline. We also utilized the Anthropic credits we were given and used Claude Code to aid us in debugging and properly mapping out our project.
Challenges we ran into
Challenges: Parsing the JSON data and putting it into fields that we could then freely use within a database was a difficult process, especially due to the fact that Greenhouse and Lever had different ways of formatting its data. We also ran into some trouble when figuring out what to use the groups of data for, which we found a solution for but were ultimately not able to fully implement into the website in time.
Accomplishments that we're proud of
Accomplishments: Our frontend is fully functional and displays everything we worked on in a clean and concise format, which was something we were worried about after the extensive backend work we did. We were also able to make Nemotron give valuable insights in our investors tab, which was a difficult task to accomplish due to the fact that it would give us a lot of jargon and useless statistics compared to what we were looking for.
What we learned
We learned that, without proper oversight, LLMs can output a lot of jargon and nonsense, so it is more difficult to implement one into a project than just making sure it functions. We also learned that a lot of data is good, but too much data without a proper organization method can clutter a project and hurt the user experience.
What's next for Signal Foundry
We plan on adding more companies via Greenhouse and Lever as well as other open-source job sites, as well as a way to automatically update data. We also have priority signals, such as "leader with no team opening" or "influx in openings in [country]" that we would like to add to the investor page.
Log in or sign up for Devpost to join the conversation.