HACK4HACKYARD • BUILD 2026 • ROUND 2
PitchForge
AI / From idea to investor-ready pitch
Generate a pitch. Stress-test the story. Improve it with structured feedback.
GENERATE
SCORE
REFINE
CHALLENGE Q&A
TEAM • PITCHFORGE
SARANSH · TANISHA�SIDDHI KUMAR JHA · SYEDA SHABIHA ARIFA
LIVE PROTOTYPE
SCAN TO OPEN LIVE APP
pitchforgeai-six.vercel.app
github.com/saransh1860-a11y/Pitch-forge-ai
HACK4HACKYARD • BUILD 2026 • ROUND 2
01 / 06
02 / PROBLEM & TARGET USERS
Good ideas get stuck before they become convincing pitches.
The gap is not just writing slides—it is a repeatable path from idea to investor-ready reasoning.
THE PROBLEM
Pitch creation is
fragmented.
Founders must translate an idea into a crisp story, support it with business logic, then anticipate difficult investor questions.
TODAY
Docs + generic prompts + scattered feedback
MISSING
A joined-up draft → critique → revision → practice loop
01
WHO FEELS IT?
Student founders · First-time entrepreneurs · Incubator teams
02
WHAT THEY DO NOW
Piece together templates, chatbot drafts and human opinions—often without a consistent scoring rubric.
03
WHY IT IS WORTH SOLVING
A repeatable feedback loop can help users find weak logic earlier—before a pitch competition, incubator review or investor meeting.
Evidence note: this deck makes no unsupported market-size or success-rate claim; the case is the observable workflow gap.
HACK4HACKYARD • BUILD 2026 • ROUND 2
02 / 06
03 / WHAT YOU BUILT
One workspace for pitch generation and improvement
From startup idea to a scored pitch, revision cycle and investor Q&A.
01
IDEA INPUT
Describe the startup
02
GENERATE
Create 10-slide draft
03
EVALUATE
Score + actionable critique
04
ITERATE
Revise + investor Q&A
PRE-EXISTING BASELINE
Already documented in the repo
AI-generated 10-slide pitch structure
0–100 venture-heuristic score + critiques
Investor challenge Q&A + answer evaluation
Firebase Auth / Firestore session persistence
TECH STACK
How the system is put together
UI
React 19 · TypeScript · Tailwind · Vite
API
Node.js / Express · Vercel serverless routes
AI
Google Gemini via @google/genai SDK
DATA
Firebase Authentication + Cloud Firestore
BUILD 2026 FOCUS: Demonstrating the end-to-end founder journey across the existing PitchForge AI workflow.
HACK4HACKYARD • BUILD 2026 • ROUND 2
03 / 06
04 / WORKING DEMO & RESULTS
The proof is the full loop—not a single generated paragraph.
End-to-end flow: startup idea → generated pitch → AI score and critique.
PITCHFORGE AI / DEMO FLOW
RUN LIVE
01 INPUT
Startup idea
Problem
Audience
Solution
Business model
FOUNDER BRIEF
02 GENERATE
10-slide draft
Problem · solution · market
model · team · traction
03 EVALUATE
Venture score
0–100
Score + structural
feedback
LIVE CHECKLIST
1
Submit a real idea
2
Inspect the generated deck
3
Review score + critique
WATCH THE DEMO VIDEO
Scan the QR to watch PitchForge AI turn an idea into a scored pitch workflow.
Scores are heuristic feedback—not a prediction of investor funding or startup success.
HACK4HACKYARD • BUILD 2026 • ROUND 2
04 / 06
SCAN TO WATCH
05 / INGENUITY & TECHNICAL DECISIONS
Go beyond generation: critique, revise, practice
Built to move beyond “generate once” toward critique, iteration and pitch practice.
CLOSED-LOOP IMPROVEMENT
DRAFT
SCORE
REVISE
CHALLENGE
PITCH
QUALITY
Score feedback guides revision; Q&A probes founder understanding separately from deck quality.
TECHNICAL DECISION
Bounded model fallback
Architecture notes specify an 8-second primary-model attempt, then fallback to a lighter Gemini model when the first path times out or fails.
SECURITY CHOICE
Keep Gemini calls server-side
The browser calls the API layer; the API key stays in server environment variables.
CHALLENGE
Avoid scores that reward empty answers
The Q&A flow uses meaning-based grading and a separate knowledge score—not just whether text was entered.
HACK4HACKYARD • BUILD 2026 • ROUND 2
05 / 06
06 / POTENTIAL & NEXT STEPS
Structured pitch prep for founders and incubators
A practical path from prototype to a tool founders and incubators can trust.
STUDENT FOUNDERS
Turn a project into a clearer story
FIRST-TIME FOUNDERS
Spot weak assumptions earlier
INCUBATORS
Give cohorts a consistent first-pass rubric
NEXT 3 BUILDS
01
Evaluate with a benchmark set
Compare scores and feedback against expert-reviewed sample pitches.
02
Ground claims in evidence
Add source-backed market/competitor research and clearer uncertainty labels.
03
Harden for real users
Run deployment smoke tests; improve error handling, privacy and export reliability.
BEFORE REAL-WORLD USE
Calibrate rubric against experienced reviewers
Measure failure rates and model costs
Make limitations and generated claims visible
A high score is feedback—not a promise of funding.
TRY THE PROTOTYPE https://pitchforgeai-six.vercel.app • SOURCE https://github.com/saransh1860-a11y/Pitch-forge-ai
HACK4HACKYARD • BUILD 2026 • ROUND 2
06 / 06