Machine Learning Engineer

Research, prototype, and productionize ML models for concrete business problems

Run end-to-end ML experiments: data → feature engineering → training → evaluation

Optimize models for latency and throughput (quantization, pruning, ONNX export)

Implement custom training loops in PyTorch for non-standard architectures

Collaborate with data teams to build clean, reliable feature pipelines

Write production inference code running in high-availability services

Track experiments rigorously; reproduce results from academic papers

Python: NumPy, Pandas, Scikit-learn (solid), PyTorch (primary deep learning framework)

Deep learning: CNNs, RNNs, Transformers, Attention mechanisms from scratch

Classical ML: gradient boosting (XGBoost, LightGBM), ensembles, SVMs

Math: linear algebra, probability, statistics, calculus — must be solid foundation

Feature engineering: encoding, imputation, scaling, feature selection, drift handling

Model optimization: quantization (int8/fp16), ONNX export, TensorRT

Experimentation: MLflow, W&B; statistical significance testing for results

Backpropagation: derive it manually once for deep understanding — not memorization

Bias-variance tradeoff; regularization techniques (L1, L2, dropout, batch norm)

Evaluation metrics by task type: AUROC, F1, NDCG, BLEU, ROUGE

Class imbalance handling: SMOTE, class weights, threshold optimization

Probability theory: Bayes theorem, MLE, MAP, common statistical distributions