Business Context

VoltLab, an electronics learning platform, wants to automatically label student-submitted op-amp circuit configurations as inverting, non-inverting, comparator, or buffer. The goal is to support instant feedback in a browser-based simulator used by ~20K students per week.

Dataset

You are given a tabular dataset generated from simulated op-amp circuits and annotated by electrical engineers. Each row represents one circuit instance with component values, connection topology indicators, and measured operating characteristics.

• Size: 48K circuit samples, 28 features
• Target: Multiclass label — inverting, non_inverting, comparator, buffer
• Class balance: Mildly imbalanced — 40% inverting, 30% non-inverting, 20% comparator, 10% buffer
• Missing data: ~7% missing in resistor fields due to incomplete student submissions; 3% missing measured outputs from failed simulations

Success Criteria

A good solution should achieve macro F1 >= 0.88 on a held-out test set and provide interpretable feature importance so curriculum designers can understand common failure modes.

Constraints

• Inference must complete in <20 ms per circuit in an API setting
• The model should be explainable to non-ML stakeholders
• Retraining will happen monthly as new labeled circuits arrive

Deliverables

1. Build a multiclass classification pipeline for op-amp circuit mode prediction.
2. Explain how the model captures the working principle of an op-amp through features such as feedback path, gain behavior, and saturation.
3. Handle missing values and mixed feature types correctly.
4. Evaluate with appropriate multiclass metrics and discuss error patterns.
5. Propose a production-ready training and inference approach.