Python Machine Learning & Data Science Project
Budget: ₹1,500 – ₹12,500 INR
I’m looking to turn a collection of well-structured tables—currently sitting in SQL and a few spreadsheets—into reliable, production-ready predictive models. The core of the engagement is predictive modeling and machine learning; everything else (EDA, cleansing, visual stories in Power BI) supports that goal.
Here’s what I need:
• Take the structured source data, verify its quality, then engineer features that make sense for modeling.
• Build, tune, and validate classification and/or regression models in Python (pandas, scikit-learn, XGBoost, or similar).
• Surface model insights inside Power BI so business users can explore results interactively.
Although the raw inputs are structured, some records link to image files. If incorporating basic image attributes or embeddings would boost accuracy, I’m open to it.
Preferred skill set
Python, SQL, Power BI, scikit-learn, XGBoost or LightGBM. Experience with NLP or other advanced techniques is a bonus for future phases.
Deliverables
1. Clean, documented Python notebooks (or .py scripts) with reproducible code.
2. A trained model saved in a portable format (pickle, joblib, or ONNX).
3. A Power BI report that visualizes key drivers, performance metrics, and allows what-if interactions.
4. A short hand-off guide explaining data flow, refresh steps, and how to retrain the model.
Acceptance criteria
• Model performance meets an agreed-upon metric (we’ll define during kickoff).
• All code runs end-to-end on fresh data using provided instructions.
• Power BI dashboards update correctly when new predictions are ingested.
If this aligns with your expertise, let’s discuss timelines and a sensible milestone plan.
Here’s what I need:
• Take the structured source data, verify its quality, then engineer features that make sense for modeling.
• Build, tune, and validate classification and/or regression models in Python (pandas, scikit-learn, XGBoost, or similar).
• Surface model insights inside Power BI so business users can explore results interactively.
Although the raw inputs are structured, some records link to image files. If incorporating basic image attributes or embeddings would boost accuracy, I’m open to it.
Preferred skill set
Python, SQL, Power BI, scikit-learn, XGBoost or LightGBM. Experience with NLP or other advanced techniques is a bonus for future phases.
Deliverables
1. Clean, documented Python notebooks (or .py scripts) with reproducible code.
2. A trained model saved in a portable format (pickle, joblib, or ONNX).
3. A Power BI report that visualizes key drivers, performance metrics, and allows what-if interactions.
4. A short hand-off guide explaining data flow, refresh steps, and how to retrain the model.
Acceptance criteria
• Model performance meets an agreed-upon metric (we’ll define during kickoff).
• All code runs end-to-end on fresh data using provided instructions.
• Power BI dashboards update correctly when new predictions are ingested.
If this aligns with your expertise, let’s discuss timelines and a sensible milestone plan.