Career Fair Agent

Turn career fair lines into interview invites.

Comment

Inspiration

Career fairs are high-pressure, but candidates use the exact same generic pitch for every booth. A recruiter cares about culture and hiring funnels, while a technical hiring manager cares about database bottlenecks and system architecture. We built BoothAgent to give candidates a real-time, context-aware "tactical battle card" in line, tailored to exactly who they are about to speak to.

What it does

BoothAgent is a mobile-first app designed for one-handed use in line.

Pre-Fair: Extracts engineering "Power Spikes" with concrete metrics from the candidate's resume. In-Line: Uses real-time web search for target companies and GPT-4o Vision to extract representative details from badges/LinkedIn screenshots. Pitch Generation: Creates a custom Battle Card (30-second hook, icebreaker, 3 smart questions, closing script) optimized for either a Recruiter or Technical Hiring Manager persona. Post-Booth: Generates instant under-300-character LinkedIn notes and follow-up emails with one tap.

How we built it

Frontend: Next.js (App Router), TypeScript, and Tailwind CSS. OpenAI Stack: GPT-4o for real-time web-grounded research, GPT-4o Vision for badge OCR, and GPT-4o-mini for low-latency pitch and follow-up drafting. Pipeline: A multi-stage agentic workflow leveraging Structured Outputs (JSON Schema) to seamlessly pass structured candidate and company context between skills.

Challenges we ran into

Orchestrating multi-model pipelines under real-time constraints. Waiting in line requires immediate results, so we refactored our sequential AI calls to run company research and candidate profiling in parallel, cutting latency in half.

Accomplishments that we're proud of

Building a true mobile-first, one-handed UI that requires zero typing while waiting in line. Achieving a clear distinction in tone and technical depth between the Recruiter and Technical Hiring Manager pitch strategies. Got over hundreds of people interested in the product after first product post on Threads.

What we learned

When chaining agentic steps together, structured JSON schemas at the API level are essential for preventing failures. We also learned to design for high-stress environments where speed and minimal interaction are key.

What's next for Career Fair Agent

Hype-Up Mode: A voice-to-text pitch practice interface with real-time feedback. Offline Caching: Pre-loading profiles and company data to bypass crowded convention hall cellular dead zones.

Built With

Share this project:

Updates