Milton Training Needs Analysis Platform
Milton is an applied ML platform developed as a capstone project in collaboration with Lockheed Martin to automate training-needs analysis from job descriptions and training documentation.
What it was
Milton was built as a capstone project in collaboration with Lockheed Martin to automate the extraction and evaluation of knowledge, skills, and abilities from job descriptions and training documentation.
The project centered on a real operational problem: manual training-needs analysis is slow, subjective, and difficult to scale across roles and departments.
Industry supervision for the project was provided by Allie Munro.
Technical approach
- Used a hybrid architecture combining RoBERTa-based language understanding, spaCy processing, SVM classification, and K-means clustering.
- Built the backend around FastAPI and Docker with a microservices structure that separated extraction, scoring, and interface concerns.
- Added a feedback loop so validated user corrections could inform retraining and improve the system over time.
Outcome
This project is a strong example of machine learning engineering tied to a usable workflow, with model logic, API performance, and front-end delivery working together.