📊 PM Portfolio Agent
An AI assistant built for business stakeholders — not engineers. Ask plain-English questions about project risk, budget, and team capacity. The agent decides which data to check, chains multiple lookups when a question needs it, and answers like a business analyst would.
- Real tool-use / function-calling across multiple linked datasets
- Chains lookups automatically when a question spans risk + capacity data
- Full reasoning trace shown for every answer — not a black box
- Built for non-technical business stakeholders — no SQL or dashboard required
Every answer is grounded in real data lookups — the agent decides which tools to call per question, and chains them when a question spans both project risk and team capacity data.
🔧 Calling tool: check_resource_allocation(month="2026-07")
📊 Result: 4 of 6 at-risk projects have at least one overallocated team member. One person is stretched to 140% across three of them — flagging as the top risk driver.
⚡ Smart Meter Anomaly Agent
A genuine agentic AI demo — not a chatbot. The agent autonomously investigates a smart meter electricity dataset, reasons step-by-step about likely causes, and validates its own findings against labeled ground truth. Built on the same kind of energy data I work with at SCE.
- Real tool-use / function-calling — not a scripted demo
- Visible multi-step reasoning before any conclusion
- Self-validated against 150 labeled anomaly readings from Kaggle
- Directly tied to my SCE energy data engineering work
Validated against 150 labeled readings from the Kaggle Smart Meter Electricity Consumption dataset — real ground-truth comparison, not a self-reported number.
🔧 Calling tool: check_weather_correlation(row_index=74)
📊 Result: 100% below similar-weather average — no meteorological explanation. Flagging as Abnormal.
