Building production-focused AI systems across Legal AI, Medical AI, NLP, Computer Vision, and Agentic AI.
Instead of collecting models, I focus on building complete AI products and research systems.
| Domain | Problem | Solution |
|---|---|---|
| Legal AI | Understanding legal documents and answering legal questions | Legal RAG System |
| Healthcare AI | Early skin cancer detection | Melanoma Classification System |
| NLP | Extracting insights from app reviews | Play Store Review Analyzer |
Legal documents are lengthy, complex, and difficult to search manually.
Built a Retrieval-Augmented Generation pipeline capable of:
- Legal document ingestion
- Chunking and indexing
- Semantic search
- Context retrieval
- LLM-powered question answering
- Legal knowledge extraction
LangChain → Vector Database → Retriever → LLM → Response Generation
A foundation for an AI-powered legal assistant operating on Indian legal documents.
Early melanoma detection can significantly improve treatment outcomes.
Developed a deep learning-based diagnostic pipeline that:
- Processes dermoscopic images
- Performs melanoma classification
- Estimates prediction uncertainty
- Generates explainable outputs
- Produces exportable reports
CNN → Test-Time Augmentation → Uncertainty Estimation → Report Generation
End-to-end medical image analysis workflow focused on decision support.
Large applications receive thousands of reviews that are difficult to analyze manually.
Built an automated review intelligence platform capable of:
- Review collection from Google Play Store
- Sentiment analysis
- Aspect classification
- Category detection
- Improvement recommendation generation
- Interactive visual analytics
Reviews
↓
Preprocessing
↓
Sentiment Analysis
↓
Aspect Classification
↓
Recommendation Generation
↓
Dashboard
Transforms raw customer feedback into actionable product insights.
Machine Learning:
- Supervised Learning
- Deep Learning
- Computer Vision
- NLP
LLM Engineering:
- RAG
- LangChain
- LangGraph
- Agentic Systems
Deployment:
- Streamlit
- Firebase
- Docker
Programming:
- Python
- SQL
- Agentic AI Systems
- LangGraph Workflows
- Legal AI
- Medical AI
- Multi-Agent Architectures
- LLM Engineering
- GitHub: https://github.com/subhasishsaha
- Email: subhasishsaha.work@gmail.com
Building complete AI systems from research to deployment.