Inspiration
Prediction markets are incredibly powerful tools for aggregating human knowledge, but trading them is currently built for power users. You have to navigate complex order books, calculate implied probabilities, and manually type limit orders. This friction keeps everyday people out. We wanted to build a novel interface that completely removes the keyboard from the equation. We asked ourselves: What if trading on-chain was as easy as talking to a friend who happens to have a PhD in quantitative finance?
What it does
Polymax Aura is a seamless, voice-enabled prediction market trading terminal. Users speak naturally to our intelligent agent "Aura," asking for trending markets, checking real-time YES/NO probabilities, and instantly executing trades without a single click.
Beneath the captivating, voice-reactive 3D interface lies a rigorous quantitative engine. While the user experiences a frictionless conversation, the backend is seamlessly translating their natural language intents into structured JSON that routes directly through the Liquid SDK to execute lightning-fast trades on the Polymarket CLOB.
How we built it
The Interface: We built a stunning, modern frontend using HTML, CSS, and Three.js for the voice-reactive orb. The Voice Layer: We integrated VAPI and Deepgram for ultra-low latency, real-time speech-to-text. We used Fish Audio TTS to give Aura a distinctive, high-quality voice. The Trading Engine: The core logic is powered by a FastAPI backend. When Aura decides to make a trade, it hits our custom vapi-webhook, which interfaces directly with our liquidtrading-python SDK wrapper. The Data: We tied Aura into both the Polymarket Gamma API (for semantic market search) and the Polymarket CLOB API (for real-time orderbook pricing).
Challenges we ran into
Building a low-latency voice agent that interacts with financial APIs is notoriously difficult. Our biggest challenge was the "time-to-first-trade." Initially, routing the voice through the cloud, parsing the intent, fetching live Polymarket data, and executing via the Liquid SDK took too long for a natural conversation. We solved this by tightly coupling our FastAPI webhook with local tool functions and using ngrok to give the VAPI cloud agent direct, zero-overhead access to our Python execution environment.
Accomplishments that we're proud of
We are incredibly proud of the fact that we took a highly complex mathematical framework (mean-reverting portfolios, Wasserstein distance arbitrage, and Sinkhorn iterations) and wrapped it in an interface so simple that literally anyone can use it. We successfully created a 100% keyboard-less trading flow from market discovery to trade execution.
What we learned
We learned that the "Agentic UI" paradigm requires a massive amount of hidden structural engineering. For an AI to reliably execute trades using the Liquid SDK based purely on voice, the tool definitions and system prompts have to be incredibly strict so the LLM doesn't hallucinate order sizes or prices.
What's next for Polymax Aura
Currently, Aura executes manual trades based on user commands. In the future, we want to allow users to speak complex automated strategies into existence. Imagine saying: "Aura, monitor the Presidential Election market, and if YES drops below 45 cents, automatically scoop up 1,000 shares via Liquid." We want to turn Aura from a voice-broker into a fully automated voice-quant.
Log in or sign up for Devpost to join the conversation.