Inspiration
Hiring is broken. I've seen it firsthand, recruiters spending hours screening candidates, scheduling interviews, and writing feedback, only to make inconsistent decisions because they're exhausted by candidate 10. I wanted to fix that. When I saw Amazon Nova's multimodal and reasoning capabilities, I thought, what if Nova could do the heavy lifting and give recruiters their time back?
What it does
HireIQ automates the entire early-stage interview process:
- Candidate uploads their resume PDF
- Nova reads it and generates 6 personalised interview questions specific to that candidate's background
- Candidate answers through a chat interface
- Nova scores every answer using extended thinking, producing a full scorecard with a visible reasoning chain, not just a number
- Recruiters see all candidates ranked and shortlisted with hiring insights
How we built it
Built on Amazon Bedrock with Nova 2 Lite doing three distinct jobs, multimodal PDF parsing, stateful interview conversation, and extended thinking evaluation. Python FastAPI handles the backend with in-memory session management. React + Vite on the frontend with three core screens: interview chat, scorecard, and recruiter dashboard. The whole thing runs locally with no complex infrastructure needed.
Challenges we ran into
Getting Nova's extended thinking to return consistent JSON was the biggest fight. The reasoning trace and the structured scorecard come back in separate content blocks and parsing them reliably took several iterations. Also structuring the prompt so Nova generates questions that are genuinely specific to a resume, not generic filler, required a lot of prompt tuning.
Accomplishments that we're proud of
The reasoning chain display. When you open a candidate's scorecard and see exactly how Nova thought through every answer, what evidence it used, why it gave that score, it feels like having a senior hiring manager explain their decision out loud. That's the moment that made this feel real.
What we learned
Nova's extended thinking is genuinely powerful but you have to design your prompts around it, not bolt it on. The reasoning trace isn't just a feature, when it's visible to the user it completely changes how much they trust the output. Transparency matters in hiring more than almost any other domain.
What's next for HireIQ
Voice interviews using Nova 2 Sonic, so candidates can speak their answers naturally instead of typing. Multi-role support so one recruiter can run parallel pipelines for different positions. And bias detection built into the scoring layer, flagging when evaluation criteria might be leading to unfair outcomes. The foundation is solid, this is just the beginning.
Log in or sign up for Devpost to join the conversation.