Intelligent annotation
Project Title: Scarecrow – Intelligent Annotation for Aircraft Engine Management
Client: Rolls-Royce (via Deloitte)
Timeline: February 2021 – January 2023
Technologies: Python, Streaming Machine Learning, Databricks
Team Size: 7–10 (Data Scientists and Data Engineers)
Recognition: Presented at AI/ML Systems Conference; appreciated by CEO & CTO of R2Factory
Publications: Presented Scarecrow - Intelligent Annotation platform for Engine Health Management in AI ML Systems conference
White papers: Demonstrating online learning on Rolls-Royce blogs
Problem Statement¶
Subject-matter experts (SMEs) at Rolls-Royce were manually reviewing sensor data from aircraft engines to detect performance issues and potential failures. This process was time-consuming, subjective, and difficult to scale across thousands of engines.
Solution¶
A “Human with AI” platform named Scarecrow was developed to assist SMEs in identifying engine anomalies. The system continuously learned from SME decisions using streaming machine learning, enabling real-time adaptation and annotation. A web-based framework was built to monitor SME interactions and train models that could predict failures in specific engine components. These models were deployed to predict failure of specific parts in an aircraft engine across over 1000 aircraft engines, helping prioritize engines that required attention.
Impact¶
The solution reduced false positives in preventive maintenance by an estimated 15%, saving over 1,200 man-hours. It enhanced the scalability and consistency of engine health diagnostics. The project demonstrated the power of combining human expertise with adaptive AI in high-stakes engineering environments.