Build a Trading Algo with AI | No Coding Required

Summary

Karthik Rangappa and Akshay demonstrate how a non-programmer can use an AI coding assistant to build a semi-automated trading signal dashboard connected to Zerodha Kite Connect. The example strategy scans the Nifty 100 using daily data for 6-day and 30-day simple moving average (SMA) crossovers. It labels bullish and bearish crossovers and ranks signals by how recently they occurred. The presenters verify an Asian Paints bearish signal on a Kite chart.

The dashboard displays signals; it does not place trades. Users decide whether to act and place orders manually. The video does not report strategy returns, risk-adjusted performance, or backtest results, and it does not establish that the crossover strategy is profitable.

Trading Ideas and Examples

• Demonstration strategy: Scan Nifty 100 constituents for bullish or bearish crossovers of the 6-day SMA and 30-day SMA, prioritizing the most recent crossover dates.
• Chart verification: An Asian Paints (ASIANPAINT) bearish crossover signal dated June 30 was checked against a daily chart using 6- and 30-day moving averages. The year is not specified in the subtitles.
• Illustrative rule examples—not recommendations:
  • Buy 200 shares of Infosys (INFY) when a moving-average crossover occurs.
  • Buy 100 shares of Reliance Industries (RELIANCE) if “HDFC” crosses above its 200-day moving average. This is presented as an example of an algo rule, not as a market view.
  • Use RSI overbought or oversold levels to generate long or short signals.
  • Calculate stock returns and standard deviations, then short a return at +3 standard deviations and go long at −3 standard deviations.
• Other possible strategies mentioned include volume breakouts, range breakouts, candlestick patterns, and scans on intraday timeframes. The presenters say similar logic could be applied to commodities if the relevant market segment is enabled.

Framework Demonstrated

1. Define the trading rules in natural language, including the universe, timeframe, indicators, signal conditions, ranking, and desired output.
2. Set up the development environment: Use Codex to install Python and Node.js, along with the Kite Connect Python SDK.
3. Create a Kite Connect app in Zerodha’s developer portal. The demonstration uses a Connect app to access historical and live market data.
4. Authenticate the app using its API key and secret, then obtain a request token and access token.
5. Build the signal engine: Retrieve Nifty 100 constituents and historical daily candles, calculate SMA 6 and SMA 30, detect bullish and bearish crossovers, and rank results by recency.
6. Create a dashboard: The demo uses a Python backend and a React/Vite front end. The dashboard includes login, account details, and a signals tab.
7. Review signals and place trades manually. The dashboard can be run to refresh the scan; execution remains with the user.

Costs, Timelines, and Operational Cautions

• The demo’s Kite Connect app costs 500 credits, equivalent to ₹500, for 30 days. The presenters say it provides access to historical and live market data; historical data may extend back to a stock’s inception.
• The request token lasts only a few minutes, so it must be used promptly. The access token keeps the session active for 24 hours.
• Codex is described as having free and paid versions. The presenter uses a paid Pro version and suggests a paid plan for ongoing, heavier development, but no subscription price is given.
• API keys and secrets are sensitive credentials. The demo saves them in a text file for convenience, while noting that a more robust approach is to use an environment file.
• The presenters stress that AI can generate the code, but users should provide detailed prompts and understand the trading rules they want implemented. The video does not demonstrate a backtest, live deployment, or profitability assessment.

Disclosure

No explicit “not financial advice” disclaimer appears in the subtitles. The content is an educational build walkthrough, not evidence that the example strategy will perform well.

Presenters and Sources

Karthik Rangappa and Akshay; Zerodha Kite Connect and its developer portal; Codex, described in the video as OpenAI’s coding agent.