About the Project
Fuel card sales in the U.S. (all sales are conducted within the United States).
Project launch: March 2024.
Part of a logistics group: The project is a division of a U.S. trucking logistics group, which is the market leader in Uzbekistan.
The company is a registered IT Park resident with offices in Tashkent (two offices), Chicago, and Orlando.
Purpose of the Role
The main goal of this role is to design and implement a set of risk-based pricing models that determine individual fuel discounts ($/gallon) for customers based on 20–30 financial, behavioral, and industry-related factors. Models should cover new, existing, and churn-risk clients, with a clear business impact evaluation.
Key Responsibilities
-
Analyze and clean large historical datasets (2–3 GB in Excel format).
-
Design and implement multiple pricing models tailored to different client categories.
-
Perform feature engineering and variable selection (20–30 features: finance, behavior, industry, etc.).
-
Train and calibrate models using algorithms such as LightGBM, XGBoost, Logistic Regression.
-
Build explainable models with SHAP, feature importance, and other interpretability tools.
-
Develop a framework for business-effect evaluation (uplift, sensitivity analysis).
-
Prepare models for use by the finance department and potential automation via API.
-
Document hypotheses, model logic, feature selection, and interpretations.
-
Provide recommendations for deployment (batch scoring, API integration, model updating).
-
Plan quarterly model recalibration and monitoring.
Requirements
-
3–5+ years of hands-on experience in Data Science or Applied Machine Learning.
-
Proven expertise in scoring, risk, or pricing models.
-
Strong Python skills (pandas, scikit-learn, XGBoost/LightGBM).
-
Experience in feature engineering and explainable modeling (e.g., SHAP).
-
Understanding of pricing logic, discounting mechanisms, and sensitivity analysis.
-
Ability to work with large Excel datasets and extract insights.
-
Strong independence in managing the full cycle: from analysis to implementation recommendations.
Nice to Have
-
Background in fintech, e-commerce, or dynamic pricing systems.
-
Experience deploying ML models (FastAPI, Docker, MLflow).
-
Knowledge of scorecard model development.
-
Experience with visualization tools (Plotly, Streamlit).
Technologies & Tools
-
Python (pandas, scikit-learn, XGBoost, LightGBM, SHAP)
-
Excel, Jupyter, SQL (optional)
-
MLflow, Streamlit (when needed)
-
FastAPI (for production deployment if required)
What We Offer
-
Competitive compensation (discussed individually based on competencies).
-
Direct access to company leadership – your expertise and ideas will be valued.
-
5/2 schedule following the U.S. production calendar for holidays and weekends.
-
Working hours: 18:00–02:00 (Tashkent time).
-
Office-based position in Tashkent.