Signl

Inspiration

We wanted to build something genuinely autonomous instead of another AI assistant where humans still make every decision. Prediction markets felt like the perfect environment to test whether an AI agent could actually reason, take action, and potentially make profitable decisions on its own.

What it does

Signl is an autonomous AI trading agent for Kalshi prediction markets. It monitors live markets, reads recent news, analyzes sentiment using NVIDIA Nemotron, estimates probabilities, manages risk, and either executes trades automatically or sends recommendations to the user through Telegram.

How we built it

We built Signl using OpenClaw via Brev as the autonomous agent framework, NVIDIA Nemotron for reasoning and sentiment analysis, Kalshi’s API for market data and trading, and Telegram as the user interface. We also built a custom watchlist system, SQLite trade database, automated scanning pipeline, and Kelly Criterion-based risk management layer.

Challenges we ran into

The biggest challenge was making the system reliable end-to-end. We had to connect live news analysis, AI reasoning, market APIs, Telegram messaging, and automated trade execution into one workflow that could continuously run without breaking. Another challenge was balancing autonomy with safety so the agent wouldn’t take reckless actions.

Accomplishments that we're proud of

We’re proud that Signl is more than just a chatbot, and that it’s a fully autonomous workflow that can monitor markets, analyze information, and make trading decisions with minimal human input. We also successfully integrated strict execution policies to securely limit what the agent can access and execute.

What we learned

We learned that autonomous agents become much harder once real APIs, state management, and continuous execution are involved. We also learned a lot about building with NVIDIA’s AI stack, especially how to use Nemotron for structured probabilistic reasoning instead of generic chatbot responses, and how important secure execution boundaries are when AI agents interact with real-world systems.

What's next for Signl

Next, we want to improve the trading strategies, add more advanced market analysis, support multi-agent collaboration, and test Signl on larger sets of live markets. We also want to explore whether autonomous agents like this can consistently outperform us and our manual trading over longer periods of time.