.png&w=828&q=75)
We have been building a one-size-fits-all orchestration agent focused on solving the problem of agentic deep research. There is an undeniable limit of context size and speed when it comes to deep search, but the need for it is very much there. The solution? A flexible, modular and lightweight architecture which delegates tasks based on recursive knowledge gathering. All your agents do what they do best, our all-in-one orchestrator guides them to unlock their maximum potential. Separation of concern, targeted data enrichment and a powerful context manager means a simple LinkedIn scraper can become a cutting edge lead generator, candidate sourcer or job tracker in a matter of minutes. Introducing a orchestration-type agent to Coral will allow users to unlock a wider range of applications. We see the landscape of Coral Agents to grow beyond the traditional capabilities of linear agents, we want part. For this hackathon, we put our orchestrator to the test with a specific application close to our hearts. We are painfully aware of the issues in recruitment today: tens of thousands of applications for any given role. Truth is, it's impossible to properly check if the role fits for every one of them. This is terrible for both the recruiters and applicants. With the All Aboard agent, we have been able to lower the cost and time taken for applicants to receive the review they deserve.
21 Sep 2025

Have you heard of Soham? In a world where hiring fraud is harder to detect, how can you identify bad actors quickly and affordably? We’re team Le Commit, and we built Unmask for the Vultr track after seeing how easy it’s become to fake resumes, reuse identities, and vibecode technical interviews. Fast-moving teams don’t have time for slow, outdated background checks, and the cost of a bad hire is too high. Unmask is an AI-powered credibility checker that helps hiring managers verify candidate authenticity by detecting inconsistencies across CVs, LinkedIn, GitHub, reference calls, and live interviews. It flags timeline mismatches, missing signals, and potential identity fraud. Reference calls are automated using voice APIs, with transcripts scored and cross-checked against prior claims. During interviews, Unmask provides real-time prompts to validate risky areas. The dashboard surfaces credibility scores, red/yellow flags, and suggested follow-ups, giving teams signal before they waste time. Built with LLaMA models, Groq API, React, Tailwind, and deployed on Vultr using a modular agent pipeline. Like at http://unmask.click/ Repo https://github.com/mousberg/le-commit
8 Jul 2025