- Processed and feature-engineered ~1M OHLCV rows across 500 Nifty 500 stocks
- Built an 8-layer modular ML pipeline
- Implemented XGBoost and Scikit-learn models
- Worked with time-series analysis, ranking and backtesting
- Added 29 automated tests
- Identified and fixed 3 production bugs
- Built an interactive Streamlit + Plotly dashboard
- Built a Retrieval-Augmented Generation (RAG) chatbot
- Implemented semantic retrieval using embeddings
- Connected custom knowledge sources with LLM responses
- Implemented conversational memory
- Built context-aware question answering
- Built a real-time digit recognition system
- Applied computer vision and image processing
- Implemented real-time prediction
- Built an NLP classification system for toxic and abusive text
- Applied text preprocessing and TF-IDF
- Worked with imbalanced datasets
- Benchmarked classification models
- Deployed the model using Flask
- Built a Python-based AI voice assistant
- Automated 15+ commands
- Integrated speech recognition and text-to-speech
- Added email automation using SMTP
- Published in IJARESM
