Inspiration
Founders pitch with optimism. Investors listen with skepticism.
We were inspired by the gap between what founders think they’re saying and what investors actually hear. That gap often appears in silence, after the pitch ends and the real evaluation begins. Most founders only discover it when the stakes are already high and feedback is limited.
SoSharks was built to surface that moment early. It lets founders rehearse their pitch against investor logic, not applause, and see their story the way sharks do, before entering the tank.
What it does
SoSharks is an AI-powered pitch and investor feedback generator.
Founders submit a startup idea or pitch, and SoSharks simulates feedback from multiple investor personas, such as angels, venture capitalists, and skeptical sharks. Each persona evaluates the pitch from a different perspective and highlights strengths, risks, and missing signals.
The result is a consolidated Investor Reality Report that shows founders what works, what raises red flags, and what questions are likely to stop the room.
How we built it
SoSharks is built as a multi-agent evaluation system.
Each investor persona operates as a distinct reasoning agent with its own priorities, such as market size, defensibility, execution risk, or credibility. The same pitch is evaluated independently by each persona, and their outputs are synthesized into a unified response.
Internally, we treat pitch evaluation as a multi-objective reasoning problem:
Investability
= f(Clarity, Market Fit, Risk, Credibility)
Rather than producing a single score, SoSharks preserves perspective diversity to reflect how real investment decisions are made.
Challenges we ran into
One of the biggest challenges was avoiding generic AI feedback. Investor responses needed to feel opinionated, realistic, and distinct, not polite or repetitive.
We also had to carefully balance tone. The feedback needed to be honest without being discouraging, and critical without being dismissive. Designing believable investor personas under hackathon time constraints required rapid iteration and ruthless prioritization.
Accomplishments that we’re proud of
We successfully built a working multi-persona investor simulation that delivers differentiated, actionable feedback from a single pitch input.
We’re proud of creating an experience that feels both playful and serious, capturing the emotional tension of pitching while providing real value. Most importantly, we built something founders can actually use to prepare for high-stakes conversations.
What we learned
We learned that founders and investors often speak the same language but interpret it very differently.
We also learned that AI is most powerful when it reveals perspective rather than offering generic advice. Designing agent systems that feel grounded, opinionated, and helpful requires thoughtful constraint, not just model capability.
What’s next for SoSharks
Next, we plan to expand SoSharks with more specialized investor personas, deeper pitch diagnostics, and iterative pitch rewriting based on feedback. Longer-term, we envision SoSharks as a rehearsal space for founders, where ideas can be tested, refined, and strengthened before entering real rooms with real stakes.
Built With
- cloud
- express.js
- github
- gradient-ai-api
- javascript
- large-language-models-(llms)
- multi-agent-ai-systems
- node.js
- postgresql
- prompt-engineering
- react
- rest-apis
- shadcn/ui
- supabase-(auth-&-database)
- tailwind-css
- typescript
- vercel
- vite
Log in or sign up for Devpost to join the conversation.