FlowFi
SocialFi risk management dashboard for creators.
Demo
About
The problem it solves Creators live in a weird in-between space financially. Their income depends heavily on engagement, views, likes, comments, but their savings or treasury often sits in volatile assets like crypto. When engagement drops, income becomes uncertain. But nothing signals them to reduce financial risk at that moment. And when engagement is strong and stable, they might be sitting entirely in stablecoins doing nothing. Right now, there’s no connection between a creator’s social performance and their capital allocation. Analytics dashboards tell you what happened. Wallet apps let you swap tokens. But there’s no bridge between the two. FlowFi connects those dots. Use of AI tools and agents FlowFi does not rely on AI agents or machine learning models in its core decision-making. We intentionally kept the risk logic deterministic and transparent. Engagement volatility and trend direction are calculated directly from Instagram media data using straightforward statistical methods. The hedge recommendation is rule-based (e.g., higher volatility + downward trend → suggest ETH → USDC). We made this choice on purpose. For financial decisions, especially during a hackathon, we felt it was more responsible to keep the logic explainable rather than hide it behind a black-box model.
How it's made
Challenges we ran into Instagram OAuth was more annoying than expected. If the redirect URI was even slightly off, the whole flow would fail and not give a helpful error. We had to carefully match the callback URL in the Meta dashboard and our local environment. On the Uniswap side, the swap flow isn’t just “press button → swap.” We had to: - fetch a quote - handle permit data - pass the correct calldata - make sure amounts were in base units - then send everything through wagmi/viem correctly If one value was wrong, the transaction would fail. Most of our time went into making sure the quote → swap → wallet signature pipeline actually worked end-to-end. Another challenge was figuring out how to turn engagement into something that feels actionable but not over-engineered. We kept it simple: measure volatility and trend, then map that to a risk level. Getting that balance right took iteration.