PitchFight AI: Practice the Pitch Before the Real Room
They lose because the first hard question comes too late.
“What is your moat?” “Who exactly pays?” “Why now?” “What happens if a bigger player copies this?”
PitchFight AI is built for the moment before that room.
It is an AI founder pressure arena where student builders can practice startup pitches, face realistic judge-style questions, enter a deal-style pressure round, and leave with a scorecard that shows exactly what worked — and what needs fixing.
⚔️ Enter the Arena — Try PitchFight AI
Why I built it
Hackathons move fast.
Students spend hours building the product, but often only minutes practicing how to explain it, defend it, and sell it.
A good demo can fall apart when the judge asks:
- who is this really for?
- what proof do you have?
- why is AI needed here?
- what is your traction?
- what is your ask?
PitchFight AI gives founders a private place to face that pressure before the real pitch.
Not as a normal chatbot.
As an arena.
What the app does
You start with a raw startup idea. PitchFight AI turns it into a structured founder briefing: problem, solution, target users, traction, competitors, and ask.
Then you choose who you want to face:
- Skeptical VC — attacks market, moat, revenue, and defensibility
- Technical Judge — questions architecture, feasibility, and AI necessity
- Hackathon Judge — focuses on novelty, demo clarity, and MVP strength
You can also choose the pressure level: Practice Mode, Judge Mode, or Investor Mode.
Once the battle starts, the AI judge asks follow-up questions across multiple rounds. You defend your idea, answer under pressure, enter a deal-style round, and finally get a scorecard showing what landed and what needs work.
How it is built
PitchFight AI runs on Hugging Face Spaces as a Gradio app, but the interface is fully custom-built instead of using default Gradio components.
The frontend is designed like a pitch battle screen: opponent cards, confidence meter, round flow, judge attacks, voice mode, deal phase, and scorecard feedback.
The backend handles:
- pitch structuring
- judge persona logic
- multi-round battle state
- voice mode
- deal round flow
- final scoring and coaching
The AI judge reasoning is powered by NVIDIA Nemotron through the backend API.
NVIDIA Nemotron usage
PitchFight AI uses NVIDIA Nemotron as the core judge and reasoning model.
The backend calls:
https://integrate.api.nvidia.com/v1
Primary model:
nvidia/nemotron-3-nano-omni-30b-a3b-reasoning
Nemotron is used for founder brief structuring, judge reasoning, pitch battle follow-up questions, deal round pressure, and final coaching feedback.
The API key is stored only as a Hugging Face Space secret and is never exposed to the frontend.
What made it work
The main product decision was to stop building “pitch feedback” and build pitch pressure.
Generic feedback says:
“Your idea could be clearer.”
PitchFight AI asks:
“How many hours do students actually waste finding groups and notes, and what evidence do you have that this is a real problem?”
That difference matters.
Founders do not just need encouragement. They need to practice defending their assumptions.
Built for Build Small
PitchFight AI was built for the Hugging Face Build Small Hackathon.
It is submitted for:
- Backyard AI — a practical tool for student founders
- NVIDIA Nemotron Quest — judge reasoning powered by NVIDIA Nemotron
- Off Brand — custom cinematic frontend, not default Gradio UI
- Best Demo — live Space, video, and write-up
- Field Notes — build decisions documented through the README and notes
Try it
Live Space: https://huggingface.co/spaces/build-small-hackathon/PITCHFIGHT_AI Demo Video: https://www.youtube.com/watch?v=s4_BzIBhqxc
If your pitch can survive the arena, it has a better chance in the room.
