xPool | Devpost

Inspiration

We came up with this idea after speaking with a recruiter about xAI's talent pipeline. We found that AI could assist in two key areas: sourcing great talent and managing candidates throughout the pipeline.

What it does

xPool uses GitHub and X to find relevant candidates and creates a comprehensive breakdown of each candidate's background. Recruiters and hiring managers create job postings with detailed descriptions that power the search. The platform highlights potential strengths and weaknesses in candidates' backgrounds, along with notable achievements and a professional summary. Recruiters can reach out to candidates and update their status directly from our dashboard. They can also flag standout elements of a candidate's profile to improve future recommendations. All functionality is powered by a Grok chat agent that recruiters interact with, while the dashboard displays candidate data and job postings.

How we built it

We built the frontend using Next.js and React. The backend uses FastAPI and Postgres for CRUD functionality. We leverage Celery for task orchestration, enabling our Grok agent to handle multi-step workflows. We use the X & Github API's for sourcing candidate profiles and tweets, and the xAI API to power our Grok agent (using Grok 4.1 Fast) for candidate analysis and recruiter interactions. We specifically used the new Grok Collection API, which handled vector embeddings of user data for semantic searching capabilities that power our natural language based querying.

Challenges we ran into

Integrating the xAI Collections API with Grok proved challenging. Filtering quality candidates from the initial pool was difficult due to noise, so we shifted to GitHub-first sourcing since the bot-to-human ratio is much lower and there's more concrete evidence to validate legitimacy—we then backfill with relevant X profile data. Customizing the Grok agent and building functionality for recruiters to add comments and highlights to strengthen future sourcing recommendations required significant iteration. Finding features that set us apart from existing recruiting tools also required careful thought.

Accomplishments that we're proud of

We successfully sourced and ranked over 1,500 candidates. We integrated data across multiple sources including X and GitHub to provide cohesive candidate summaries. We also made it much easier for recruiters to access the candidate pool by scraping contact information from different sources.

What we learned

Sourcing great talent is genuinely hard. It requires synthesizing multiple signals from a candidate's profile and converging on a decision amid overwhelming data. This is where AI shines—particularly in screening out weaker candidates with lots of low-quality work to sift through, and in surfacing unique strengths that might otherwise be missed.

What's next for xPool

We plan to integrate with existing recruiting tools for seamless data management, including interview notes and other metadata within candidate summaries. We'll fine-tune semantic search with a domain-specific embedding model for improved accuracy. We also want to expand sourcing to additional platforms like Codeforces, HackerRank, and Devpost.