AllProjects

Production-ready multi-agent clinical decision system using Microsoft Agent Framework on Azure AI Foundry. Five specialized agents for safety analysis, contraindication evaluation, dosing analysis, and multilingual patient counselling. Generates comprehensive medication safety reports in under 90 seconds. HIPAA-compliant with 100+ language support and 100% evaluation pass rate.

Production-grade recommendation system processing 6.7M Amazon reviews across 31,100 products. Achieves 85.4% NDCG@10 using hybrid retrieval — BM25 + dense BGE embeddings + cross-encoder reranker — with sub-1s latency. Deployed on GCP with Qdrant Cloud vector DB.

Production RAG system for financial Q&A over 6.7M historical stock prices and news articles using LlamaIndex, LangChain and Google Vertex AI embeddings. Full CI/CD with DVC versioning, Docker orchestration, and Slack alerting.

End-to-end image classification pipeline on 101 food categories using EfficientNetV2B0 with multi-stage training strategy (feature extraction → fine-tuning). Mixed-precision compute, TensorBoard tracking, 90%+ validation accuracy.

Deep learning pipeline on CHB-MIT dataset using CNN and LSTM architectures with Welch PSD and STFT spectral analysis for pediatric epilepsy detection. Comparative analysis across CNN, LSTM, and Decision Tree classifiers.

End-to-end customer segmentation on 500K+ transactions across 37 countries using RFM methodology and KMeans clustering (k=3). Identified 35-hour repeat purchase cycle, 2.27% return rate, and geographic revenue patterns.