ChatWithWiki AzureML RAG Demo
Azure ML Prompt Flow retrieval-augmented generation demo using Jinja prompt templates, retrieval orchestration, and public wiki-style content.
Azure MLPrompt FlowPythonJinja
Research Archive & Data Science Foundations
Selected experiments that show the foundation behind current AI systems work.
Healthcare analytics, HEDIS/RAG concepts, visualization notes, GPU planning, forecasting, and reusable ML patterns are curated here as background evidence for the flagship systems.
Curation principle
Foundation work is useful when it is labeled clearly.
These repos show analytical breadth, research habits, and early workflow patterns. Medication recommendation content remains explicitly non-clinical.
Azure LOS Prediction
Length-of-stay regression work using a Microsoft toy hospital dataset, Azure Notebooks, and baseline regression methods.
PythonAzure NotebooksRegressionRandom Forest
RAG Implementation for HEDIS SSD
Healthcare quality RAG concept showing de-identification pseudocode, clinical-note preprocessing ideas, and HEDIS SSD retrieval workflow shape.
PythonspaCyRAG conceptHEDIS
HEDIS Dashboard
Early Streamlit dashboard scaffold that shows healthcare analytics UI foundations.
PythonStreamlit
Medication Recommendation Experiment
LangChain medication-recommendation concept retained as an experimental healthcare LLM example with explicit non-clinical boundaries.
PythonLangChainLLM concept
HNSC Activation Visualization
t-SNE visualization concept for final-layer activations with slide images.
Pythont-SNENumPyscikit-learn
GPU Analysis for Healthcare AI Workloads
GPU comparison notes for healthcare AI workload planning across training, inference, medical imaging, agentic systems, and multi-model orchestration.
GPU planningHealthcare AI workloadsInferenceTraining
ML Templates
Machine-learning templates repository covering reusable preprocessing, evaluation, model selection, and visualization patterns.
PythonML templatesEvaluationVisualization
Safety
No clinical or sensitive-data claims are made for lab projects.
The lab uses toy data, concept repos, local demos, and research notes to show technical foundations without sensitive records or private workflows.