All Projects
Agentic AICase Study·97% Match

Disaster Response AI Copilot

Agentic AI copilot for real-time emergency response. Built on Azure OpenAI (GPT-4o) and Azure AI Foundry with a Retrieval-Augmented Generation pipeline backed by Azure Cognitive Search. Streamlit UI provides situational-awareness dashboards and actionable decision support during natural disasters. Python data pipeline in data_pipeline.py feeds live feeds into the RAG index.

GitHub ↗
PythonAzure OpenAIAzure AI FoundryRAGAzure Cognitive SearchStreamlitMachine Learning
Production30sDecision latency91%Retrieval accuracy12+Data sources4%Hallucination rate★ Featured
30s
Decision latency
91%
Retrieval accuracy
12+
Data sources
4%
Hallucination rate
🔴 The Problem
▸

Emergency teams had 8+ minute decision cycles with fragmented, siloed data sources

▸

No unified AI layer to prioritize life-saving actions across real-time feeds

▸

High cognitive load on responders during critical surge events

✅ The Solution
▸

Hybrid RAG combining semantic + BM25 retrieval across 12+ live data sources

▸

Natural language query interface — no training needed for field responders

▸

Source-cited answers prevent hallucination in high-stakes context

📈 Impact & Results
▸

Decision latency: 8 min → 30 seconds

▸

Hallucination rate dropped from 18% to 4%

▸

Deployed on AWS ECS with auto-scaling

▸

91% retrieval accuracy on disaster domain queries

Full Tech Stack
PythonAzure OpenAIAzure AI FoundryRAGAzure Cognitive SearchStreamlitMachine Learning

More Projects

Machine Learning
Crypto Trade Prediction
Machine Learning
Heart Disease Prediction with Generative AI
Machine Learning
Car Price Prediction — MLflow + Kafka + Debezium
Machine Learning
Movie Recommendation — Cosine Similarity & Collaborative Filtering
Interested in working together?
Let's build something impactful.