AI Stock Day-Trading Sentiment
Budget: ₹1,500 – ₹12,500 INR
I’m building a day-trading system that works exclusively with US-listed stocks and bases every decision on live market sentiment. Price prediction and automated execution can come later; right now I want a solid sentiment-driven engine that issues high-frequency buy/sell signals I can plug into my own trading workflow.
Here’s what I need from you:
• A sentiment analysis model (NLP) that ingests real-time news headlines, social-media posts, and company filings, converts them into a normalized sentiment score, and refreshes fast enough for intra-day decisions.
• A signal-generation layer that combines those sentiment scores with basic technical indicators (e.g., VWAP, momentum) to produce clear entry and exit recommendations for the current trading session.
• A back-testing framework so I can replay at least the last three years of intraday data and evaluate win-rate, Sharpe, drawdown, and latency.
• Clean, well-commented Python code (PyTorch, scikit-learn, or TensorFlow—your choice) plus setup instructions so I can run everything on my local machine or an AWS instance.
The final delivery is a Git repository with reproducible notebooks/scripts, sample configuration files, and a short README outlining how to connect to data feeds (Polygon, Finnhub, Twitter, etc.) and how to trigger the back-test.
If you’ve built real-time NLP pipelines for trading before, or have examples of sentiment models beating baseline technical strategies in intraday stock markets, I’d love to see them.
Here’s what I need from you:
• A sentiment analysis model (NLP) that ingests real-time news headlines, social-media posts, and company filings, converts them into a normalized sentiment score, and refreshes fast enough for intra-day decisions.
• A signal-generation layer that combines those sentiment scores with basic technical indicators (e.g., VWAP, momentum) to produce clear entry and exit recommendations for the current trading session.
• A back-testing framework so I can replay at least the last three years of intraday data and evaluate win-rate, Sharpe, drawdown, and latency.
• Clean, well-commented Python code (PyTorch, scikit-learn, or TensorFlow—your choice) plus setup instructions so I can run everything on my local machine or an AWS instance.
The final delivery is a Git repository with reproducible notebooks/scripts, sample configuration files, and a short README outlining how to connect to data feeds (Polygon, Finnhub, Twitter, etc.) and how to trigger the back-test.
If you’ve built real-time NLP pipelines for trading before, or have examples of sentiment models beating baseline technical strategies in intraday stock markets, I’d love to see them.