Keryx | Devpost

Inspiration

The financial world is fragmented. To get a complete picture of your wealth, you currently need a banking app for cash, a brokerage for stocks, a wallet for crypto, and a news terminal to understand market movements. We realized that while data is abundant, intelligence is scarce.

At the same time, AI models are advancing at an exponential rate. They are no longer just chatbots; they are becoming capable reasoning engines that can analyze complex economic scenarios better than human analysts. We were inspired to harness this power to create an automated, AI-first wealth manager. We wanted to build a "Bloomberg terminal for the everyday investor"—not just a dashboard that shows what happened, but an intelligent agent that actively helps you model what will happen.

What it does

Keryx is an AI-powered financial operating system. It aggregates real-time data across:

Banking: Transactional insights and spending habits. Markets: Live stocks (via FinnHub) and Cryptocurrencies (via CoinGecko). Macroeconomics: Inflation (CPI) and Interest Rates (via FRED).

The core of the system is the Keryx Engine and the AI Advisor, which fuses this data to generate live investment projections. Instead of static charts, users get an interactive sandbox where they can simulate strategies, all guided by an AI that understands the real-time impact of inflation and market volatility on their specific portfolio. This is all built alongside an interactive AI Advisor that is fine-tuned to your personal finance data and gives personalised recommendations, forecasts and insights.

How we built it

We built Keryx as a modern, high-performance web application:

The Core (Liminal AI): We leveraged Liminal's secure infrastructure to power our conversational AI. The "AI Assistant" login mode demonstrates how Liminal handles secure authentication while providing immediate value through conversation.

The Frontend: Built with React and TypeScript, focusing on a premium "Glassmorphism" aesthetic. We used Recharts for dynamic data visualization and Framer Motion for seamless transitions.

The Data Cloud: We engineered a robust fetching layer that synchronizes disparate data sources: CoinGecko API for real-time crypto pricing (BTC, ETH, SOL).

FinnHub API for institutional-grade stock/ETF data.

FRED API for federal economic data, giving our AI "macro-awareness."

Challenges we ran into

Our biggest challenge was state synchronization. Combining a conversational AI widget (NimChat) with a reactive dashboard required careful management to ensure the AI's "advice" perfectly matched the live graphs the user was seeing.

We also tackled complex rendering issues. At one point, a namespace conflict between our icon library and charting library caused the entire engine to crash—a bug that taught us the importance of rigorous dependency management.

Finally, integrating Financial Mathematics was successfully tricky. We had to implement real-time compound interest projections that dynamically adjust for the live inflation rate fetched from the Federal Reserve, ensuring our math was grounded in reality, not just optimism.

Accomplishments that we're proud of

True Real-Time Integration: We successfully merged three distinct public APIs (Crypto, Stocks, Economics) into a single, cohesive interface that updates live.

The "Keryx Engine" Experience: We turned a standard form into an interactive "flight deck" for finance. The way the projections update instantly as you slide the "Time Horizon" bar feels magical.

Seamless Onboarding: The dual-mode login (Standard vs. AI Assistant) creates a unique, welcoming entry point that sets the tone for the rest of the application.

What we learned

Macro Matters: By feeding inflation data into our models, we realized exactly how much "silent" wealth erosion occurs. This made our AI features more urgent and valuable.

UX is Trust: In fintech, a glitchy UI makes users fear for their money. We learned that polishing loading states and transitions is just as critical as the backend logic.

The Power of Liminal: Building on Liminal allowed us to fast-track security and privacy, letting us focus entirely on building the intelligence layer rather than reinventing auth wheels.

What's next for Keryx

We are planning to close the loop between insight and action.

Automated Execution: Allowing the AI to not just suggest portfolio rebalancing, but to execute it (with user approval).

Forex Integration: Expanding our data sources to include live foreign exchange rates for global investors.

Personalized Financial Health Score: Developing a proprietary "Keryx Score" that rates a user's financial resilience based on real-time spending and market conditions.