Sharktank AI

Sharktank AI helps founders get grilled by AI sharks before the real investors do. Today, founders rehearse slides, but investors test conversations: they interrupt, push back, and drill into the numbers - yet most founders never actually train for that, and less than 1% of startups will ever raise from top-tier funds.

Sharktank AI is a low-latency voice agent that listens to your pitch and talks back in real time, using sharp investor sense and realistic interruptions inspired by 10+ years of Shark Tank, with AI voices like Mark Cuban and other sharks.

It smashes you for weak answers on what really matters - CAC, LTV, TAM, margins, traction - so when your one real shot comes, you’re ready for the “I’m in.”

Under the hood we’re built with ElevenLabs API, anam.ai for LLM orchestration, a Next.js front end on Vercel with Clerk auth, and GitHub-powered dev flow, while bolt.new and Cursor accelerate development. N8n used for summarizing conversations and extracting insights for continuously training models. Also used for a summary email.

Our team has the right to win: 20+ years of product building, 20+ years in engineering, and 12 years in design behind this experience.

Shartank AI is a multimodal service which uses both ElevenLabs and aman.ai. We have a working Mark Cuban prototype hosted here: https://sharktank-silk.vercel.app/

Loom video: https://www.loom.com/share/a44574c4e221405492e9c9e1198c31a3

No prior work.
Team was built during the hackathon
We generated the idea and built from scratch during the hackathon