Trade$im

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built
TradeSim AI is an AI-powered paper trading and strategy backtesting platform. It helps retail traders and finance students test their trading strategies using historical data without risking real capital.

Users would pay a $10/month subscription for this because existing backtesting software is often expensive or requires complex coding. TradeSim simplifies this by combining traditional technical indicators (SMA, RSI) with an open-weight language model acting as the decision-making layer, allowing users to test logic via natural language parameters.

Code

anupam99jain / paper-trader

simulated trading platform designed to test and evaluate trading strategies using llms

Simulated trading platform designed to test and evaluate trading strategies using LLM-guided decisions.

This project is a lightweight paper trading engine that combines traditional technical indicators with large language model (LLM) decision support, allowing rapid experimentation with strategy logic without risking real capital. It continuously fetches market data, computes indicators (like SMA & RSI), prompts an LLM for trading decisions, and simulates virtual trades.

-> Features

1. Simulated order execution with virtual capital
2. Built-in technical indicators (SMA, RSI)
3. LLM-powered trade decisions (BUY/SELL/HOLD)
4. Portfolio value tracking in real time
5. Easy to extend with more indicators and strategy logic

-> How It Works

1. The script fetches current market price data periodically.
2. It computes technical indicators like 20-period SMA, 50-period SMA, and 14-period RSI.
3. The current market state is formatted into a prompt and sent to an LLM.
4. Based on the model’s (Qwen 2.5: 7b) output (BUY, SELL, or HOLD), the engine executes a…

View on GitHub

How I Built It
The backend is built in Python.

I integrated Qwen 2.5 7B as the open-weight model, running locally via Ollama. The trading loop fetches market data and calculates technical indicators (20-period SMA, 50-period SMA, 14-period RSI). This market state is injected into a prompt for the model, which outputs a strict BUY, SELL, or HOLD decision. The engine executes the virtual trade, calculates fees, and updates the simulated portfolio performance.

By keeping the architecture modular, the open-weight AI component can be easily swapped out as newer, faster models are released.

Why Does Open Innovation Matter?
For a side hustle, open innovation directly impacts unit economics. If I relied on closed APIs (like OpenAI or Anthropic), API token costs would scale with every backtest a user runs, destroying profit margins.

Using open-weight models allows me to offer unlimited backtesting to my users for a flat monthly fee. Additionally, it ensures my users' proprietary trading strategies are processed locally and remain completely private.

My First User
My friend, who originally inspired the week 1 prototype, used the platform to backtest a momentum-based crypto strategy. After seeing the dashboard track his simulated portfolio's 12% growth over a test month, he became my first paying customer by signing up for the $10/month tier to access extended historical datasets.