Inspiration
Despite the hype, the fear of AI replacing everything is kind of absurd. There’s still so much room for AI to grow—especially in the finance and insurance sectors. In fact, every executive our team has ever worked with—VPs, directors, senior managers—shared the same vision: AI isn’t a threat, it’s an opportunity. But here’s the catch—most teams don’t even know where to start when it comes to hiring AI-literate talent. Many non-technical managers want to modernize, but they lack a solid framework to evaluate candidates in such a fast-changing field. That’s why we built DevHunt: a smart recruiting platform that helps employers find top-tier tech talent based on real, verifiable hackathon achievements and AI-assisted profile matching.
What it does
DevHunt is a scouting tool that helps companies identify top hackathon talent aligned with their job needs. Users input a job description or select filters like industry, skills, and location. That input is analyzed by Gemini AI, which extracts key technical requirements—up to 50 keywords, plus category and location data. Behind the scenes, we’ve pre-scraped over 1,200 award-winning Devpost projects, along with metadata like tech stack, project descriptions, and team members. We then match user input with these projects using a compatibility scoring algorithm. From the top matches, we identify associated team members, enrich their profiles by aggregating all their hackathon history, and fetch their GitHub and LinkedIn (via real-time scraping). Finally, Gemini AI rates their technical skills and generates profile summaries—all packaged into elegant, filterable profile cards for the frontend.
How we built it
We started by building a responsive frontend using React on Replit. The user inputs are sent to a Flask backend that leverages Gemini AI to extract structured filters from natural language descriptions. Our backend then matches this data against a MongoDB collection of 1,200+ Devpost projects that we scraped in advance using MangoSQL. Each match is scored based on overlapping skills, domains, and context. We then extract developer profiles from the matched projects, aggregate their history, and scrape their GitHub/LinkedIn links if available. Finally, we use GenAI again to evaluate technical ability and generate a rating out of 10, wrapping everything into a pretty website that populates the profile cards.
Challenges we ran into
Our biggest challenge was designing a matching algorithm that balanced accuracy with efficiency—especially given our limited time and beginner coding experience. Figuring out how to iterate through large datasets without slowing the system down took a lot of brainstorming. Another challenge was defining a fair and useful judging criteria—how should we weigh skills, past wins, and project relevance? Finally, as first-time hackathon participants, we faced lots of roadblocks in setting up backend/frontend communication, scraping pipelines, and understanding how to make the entire thing come together.
Accomplishments that we're proud of
We’re proud that we were able to go from an idea to a working full-stack product in under 24 hours—especially with little prior experience. We successfully integrated AI, scraping, and a custom dataset into a live application. We also loved that we tackled a real-world problem that people in hiring actually face today. Seeing it all come together with styled profile cards and real data felt incredibly rewarding.
What we learned
We learned how to build a complete pipeline—from user input, to AI parsing, to database queries, to frontend display. We picked up Flask, React, Replit deployment, MongoDB, scraping, and Gemini AI all on the fly. We also learned how important data structuring and optimization is when building anything scalable. Perhaps most importantly, we learned how to work as a team under pressure and solve problems creatively.
What's next for DevHunt
We want to expand our dataset beyond Devpost—maybe including GitHub repositories, Kaggle submissions, or LinkedIn portfolios. We also want to improve our matching algorithm and scoring mechanism, allowing for faster queries and deeper analysis. Adding filters like “most recent win,” “collaboration frequency,” or “growth potential” could give recruiters even more insight. Finally, we hope to make the tool more customizable and user-friendly, with better UI/UX and analytics on candidate pools.
Log in or sign up for Devpost to join the conversation.