Explorar las categorías
Explora
Fiverr Pro
Español
$
USD
I will build and train custom machine learning models
Acerca de este Servicio
I design, train, and deploy machine learning models that solve real problems not just notebook experiments.
Recent modeling work:
Customer churn prediction stacked ensemble (XGBoost + LightGBM + Logistic Regression), validated against a 1M-row production-scale dataset
Customer sentiment analysis DistilBERT fine-tuned on 568K+ real reviews, 86% accuracy, F1 0.88
Multi-hazard weather risk model 7-head multi-task PyTorch network trained across 55 global cities
What I bring:
Rigorous, iterative EDA I don't skip straight to modeling
Model comparison and tuning, not just a single default model
Explainability (SHAP / feature importance) so you understand why the model predicts what it predicts
Deployment included a working API on your own hosting, not just a saved model file
Honest reporting genuine model limits get documented, not oversold
Whether you have a dataset that needs a real model built around it, or an existing model that needs improvement, explainability, or deployment I'll handle it end to end.
Message me before ordering so we can confirm your dataset and goal accurate scoping means a better result for you.
Lenguaje de programación:
Python
•
SQL
•
Colab
•
MLflow
Mi porfolio
FAQ
What kind of ML problems can you help with?
Classification, regression, clustering, time-series forecasting, image and audio preprocessing, object detection and NLP/sentiment tasks. Send me your dataset and goal and I'll confirm fit before you order.
Do I need to provide the dataset?
Yes, you will provide the data (or point me to a public source). I handle cleaning, EDA, feature engineering, modeling, and evaluation from there.
Will I understand why the model makes its predictions?
Yes. Standard and Premium include explainability (SHAP or feature importance), so predictions aren't a black box.
Will the accuracy numbers be honest, not inflated?
Always. I report real evaluation metrics and known limitations. A model you can trust matters more than one that just looks good on paper.
What if my dataset is small or messy?
That's normal, most real datasets are. I do full EDA and cleaning first, and flag upfront if data quality will limit results, before you spend the budget.
Do you handle deployment, or just hand over a model file?
Standard and Premium include deployment as a working API/app on your own free-tier hosting that you own(no recurring cost through me).
