Multi-Agent AI Compliance Platform
An AI-powered compliance tool that helps investment advisory firms review marketing materials before release. It scans emails, social posts, brochures and pitch decks, flags statements that could create regulatory risk, and links each one to the relevant firm policy with a clear risk rating and explanation. Results are compiled into an executive summary for the Chief Compliance Officer. In testing, it caught every factual claim and reached 92% overall accuracy, speeding up review while keeping final decisions with people.
Built with: Microsoft Azure AI, LangGraph, Python, and Streamlit.
Agentic Contract Risk Intelligence Platform
This AI-driven Contract Intelligence Platform combines retrieval-augmented generation (RAG) with a coordinated team of specialized AI agents to assess risks, recommend negotiation strategies, and summarize critical findings for business leaders. The solution improves contract review efficiency, supports informed decision-making, and enhances consistency across complex agreement evaluations.
AI Security Alert Triage Pipeline
End-to-end alert triage pipeline that uses Azure OpenAI (RAG over Azure AI Search) to automatically classify cloud security alerts as genuine incidents or false positives, with a human-in-the-loop review step for edge cases. Backend (FastAPI) is deployed on Azure Container Apps with automated CI/CD on merge; a Streamlit UI provides an analyst-facing review workflow. Evaluated against a labeled test set to measure precision on incident detection and false-positive suppression.
Customer Service Agent
Hope Harvest is an AI-powered customer support agent built in Microsoft Copilot Studio for the fictional nonprofit Hope Harvest. It provides instant answers to questions about donations, volunteering, eligibility, locations, hours, and community services using enterprise knowledge sources, semantic search, and natural language capabilities. Designed conversation flows, trigger phrases, escalation paths, and knowledge management to deliver accurate, user-friendly support while reducing staff workload.
.png)
.png)

.png)