In Justin Norman's latest post, he tackles a challenge familiar to many developers and data scientists: how to showcase past work when the original code is owned by a former employer. His solution? Build a simulation engine to recreate the problem space without proprietary data or IP, then reconstruct the solution using modern tools.
This approach struck me as brilliant for two reasons. First, it respects intellectual property boundaries—a must in our industry. Second, it allows you to demonstrate not just what you built, but how you’d build it now with updated frameworks and techniques. For example, Justin reimplemented a freight forecasting system using GRUs and Prophet, and a security event clustering pipeline with K-means and LSA—both reflecting current best practices.
This isn’t just about recreating code; it’s about showcasing adaptability, problem-solving, and technical growth. By rebuilding projects from the ground up, you prove you understand the fundamentals, not just the implementation details locked away in a corporate codebase.
For anyone struggling to demonstrate real-world experience in interviews or portfolio reviews, Justin’s method offers a clear, ethical path forward. Dive into his full post to see how he generated synthetic data, trained models, and even built a production-style serving layer—all from scratch.
Check out the original article and code here: Someone Else Owns My Best Code, So I Rewrote It
Follow Justin Norman's work: Bluesky | GitHub | LinkedIn | Blog
Top comments (0)