Skip to content
All work

AI & Data

IT Audit & Anomaly Detection Suite

Three ML models turned into audit-ready findings by LLM agents.

The problem

Help IT auditors decide what looks anomalous, how risky it is, and what to review first.

How I built it

  1. 01Profiled system behavior with K-Means clustering and Apriori association rules to flag anomalous clusters.
  2. 02Scored technological risk with an SGDRegressor model.
  3. 03Ordered the review queue with a simulated-annealing metaheuristic.
  4. 04Used CrewAI + Gemini agents to translate each model's output into findings and mitigation recommendations.

Outcome

  • Statistical output delivered in a format auditors can act on directly.

Next project

Sales Forecasting Model