Self-directed project
Energy and Renewable: Machine Learning for the Power Grid
Junior Machine Learning Engineer · AI / ML · August 2026
A ProoV case study · educational project, not employment
Built an end-to-end gas-turbine TEY predictor using leakage-free sensor features, with CO and NOx excluded because they are downstream emissions. Trained and compared a LinearRegression baseline and a RandomForest on a time-based held-out split, then evaluated the winner honestly with MAE, R2, and a residual read. The final model achieved about 1.53 MWh MAE and 0.983 R2 on unseen later-period data, with a clear model card describing intended use, limits, and drift monitoring.
Graded against
- Build and validate the model correctly35%
- Optimize and evaluate honestly25%
- Write the Model Card (clarity + honesty)25%
- Explain it to a non-expert15%
Passed · pass mark 60/100
What stood out6
- Correctly dropped CO and NOx because they are downstream emissions and would leak future information into the model.
- Used an earlier-period train / later-period test split instead of a random shuffle, which is the right validation pattern for ordered operational data.
- Compared a LinearRegression baseline against a RandomForest and showed a clear held-out MAE improvement.
- Noted heteroscedasticity and slight positive bias at the top end in the residual read, showing calibrated evaluation rather than overclaiming.
- Leakage-Free Feature Selection
- Time-Based Holdout Validation
The work I submitted18 tasks
10 tasks · 11 lines · Python, JavaScript · 8 written answers · 422 words
- Problem Frame238 characters
- Leakage Flags113 characters
- Train Validate BaselinePython · 1.4k characters
- Train Validate Forest943 characters
- Train Validate124 characters
- Residual Read168 characters
- Drift Monitor91 characters
- Model Card1.5k characters
- Teach BackJavaScript · 848 characters
- Jupyterlite Code2 lines
- Problem Frame18 words
- Leakage Flags1 word
- Train Validate Baseline141 words
- Train Validate Forest76 words
- Train Validate1 word
- Residual Read18 words
- Drift Monitor1 word
- Model Card166 words
Summarised on purpose — the submitted code and writing stay private so this page cannot be reused as an answer key. The full submission sits behind the verified certificate.