Eko Samuel working at his software development workstation

AI (Machine Learning) Engineer

I’m Eko Samuel, a software engineer with over ten years of experience building full-stack applications, backend systems, and cloud infrastructure. My work combines AI and machine learning with practical engineering to solve real-world problems.

Connect with Eko Samuel

All AI Projects

Case study / 01

Predictive Maintenance Using Deep Learning — NASA C-MAPSS Remaining Useful Life Prediction

Predicts aircraft-engine remaining useful life from multivariate sensor histories in the NASA C-MAPSS dataset. The project compares MLP, LSTM, and temporal convolutional networks to study degradation patterns and support maintenance planning.

  • PyTorch
  • MLP
  • LSTM
  • TCN
  • Time-series forecasting
Project evidence / Visual exploration01 / 03

01 / Context & purpose

The project in detail

Maintenance decisions depend on estimating how much useful operation remains before equipment reaches failure. This project treats engine health as a time-series regression problem, using the NASA C-MAPSS simulated turbofan degradation datasets to connect sensor observations with remaining useful life.

The repository contains MLP, LSTM, and TCN implementations, saved models, per-engine predictions, and diagnostic plots. The different architectures provide ways to compare a feed-forward baseline with models that explicitly learn temporal relationships.

02 / Engineering perspective

How the project works

  1. Prepare sensor histories and remaining-life targets, inspecting raw signals and the effects of preprocessing.

  2. Train the neural-network architectures and inspect loss curves to understand optimization behavior.

  3. Compare predictions with held-out engine lifetimes using RMSE and MAE, alongside per-engine plots that expose overestimation and underestimation.

03 / Engineering perspective

Results & Outcomes

  1. Saved TCN FD001 evaluation reports test RMSE of 37.96 cycles and MAE of 28.25 cycles.

  2. Exports per-engine predictions, training curves, and evaluation summaries for inspecting model errors.

04 / Engineering perspective

Why this project matters

  1. Shows how sensor data can support condition-based maintenance rather than relying only on fixed service intervals.

  2. Makes prediction errors visible at the engine level, where an overly optimistic lifetime estimate can have different consequences from an early maintenance recommendation.

  3. Creates a reproducible comparison of neural architectures using a public degradation benchmark.

05 / Potential next steps

Advancing AI & Future Directions

Research opportunities and proposed extensions, rather than completed achievements.

  1. Add calibrated uncertainty intervals so maintenance planners can consider confidence as well as a single lifetime estimate.

  2. Evaluate generalization across operating conditions and other C-MAPSS subsets, with explicit separation of engines across training and evaluation.

  3. Explore physics-informed learning and online drift monitoring before considering use with real equipment.

Explore the next projectTennis Performance Classification Using Label Distribution Learning and Weighted k-Nearest Neighbors