AI-Powered Trading Bot
An advanced AI-driven stock trading bot leveraging FinBERT sentiment analysis, technical indicators, and sophisticated risk management for automated intelligent trading.

The constraint
Manual trading is emotional and time-consuming. Traders need systematic approaches that combine market sentiment with technical analysis while managing risk.
The approach
Built an automated trading system combining FinBERT sentiment analysis of financial news with technical indicators (SMA, RSI). Implemented strict risk management with portfolio-wide and position-level limits.
- 1Feature design
Combined FinBERT news sentiment with technical indicators (SMA/RSI) to produce a single signal.
- 2Risk layer
Implemented portfolio-wide and per-position limits to bound downside before any order is placed.
- 3Paper trading
Validated the strategy against live data with no capital at risk before enabling execution.
- 4Execution + monitoring
Wired the Alpaca API for order placement and logged fills, signals, and risk states for review.
Outcomes
- Automated trading with sentiment + technical signals
- Real-time and paper trading modes
- Portfolio-wide risk management
- Integration with Alpaca Trading API
- Add walk-forward backtesting with transaction costs and slippage modeled honestly.
- Introduce regime detection so the strategy stands down in conditions it was not designed for.
- Track live-vs-paper divergence as a first-class metric.
Engineered an AI-powered trading bot that automates stock trading by analyzing market sentiment using FinBERT and integrating key technical indicators such as SMA and RSI. The bot employs strict risk management strategies, including portfolio-wide and individual position risk limits.
Integrated with the Alpaca Trading API, enabling real-time trading and paper trading for safe strategy testing before live deployment.
Related projects
- Read →
FraudStream
A production-grade streaming pipeline that scores financial transactions for fraud in milliseconds rather than hours, built on Kafka, Spark Structured Streaming, and a Snowflake medallion warehouse.
- Read →
Talker
An open source teaching assistant that answers course questions from the syllabus, slides and lecture materials — with citations, and YouTube timestamps when a video covers the answer.
- Read →
qsensor-sim
A physics-based quantum sensor simulation API for GPS-denied navigation research — synthetic IMU data with quantum decoherence effects, processed through Kalman filtering and bias correction.