Senior Machine Learning Engineer, Causal & Decision Systems

CSC Generation Enterprise; United States; On-site; USD 180,000 - 240,000; Full time

Job summary

CSC Generation is building closed-loop decision systems that use machine learning to operate consumer businesses more intelligently. The role focuses on estimating causal response, quantifying uncertainty, and guiding actions within guardrails, with the aim of producing economic lift in controlled experiments.

The work spans causal modeling, uncertainty estimation, policy learning, and production ML infrastructure, with emphasis on evaluating challengers and deploying safe, automated decision

Job description

CSC Generation is building closed-loop decision systems that use machine learning to operate consumer businesses more intelligently.

We are starting with pricing and expanding into areas such as inventory, purchasing, promotions, marketing, and assortment.

The Role

You will help build systems that:

**estimate causal response + quantify uncertainty → choose actions → generate useful information → observe outcomes → update policies → evaluate challengers → deploy within guardrails**

We want to answer questions such as:

• What happens **because we change a price**, rather than simply what happens next?
• How should uncertainty affect a decision?
• When should the system exploit what it knows versus experiment to learn?
• Can we estimate the value of a challenger policy before fully deploying it?
• How do we optimize economic outcomes while respecting inventory, margin, vendor, customer, and operational constraints?

What You’ll Work On

Depending on your background, you may work across:

• causal and heterogeneous treatment-effect modeling;
• uncertainty estimation and calibration;
• contextual bandits, active learning, or sequential decision-making;
• policy learning and constrained optimization;
• counterfactual and off-policy evaluation;
• experimentation and champion/challenger systems;
• production ML infrastructure, monitoring, and automated deployment.

We care about selecting the right method, not using a particular framework.

What Success Looks Like

Success is not a better offline metric.

The systems you build should produce measurable economic lift in controlled experiments, generalize across businesses, learn from their own interventions, and safely automate an increasing share of real commercial decisions.

Over time, the goal is simple:

**the system should become better at operating the business because it has operated the business.**

What We’re Looking For

We care more about exceptional technical ability and judgment than matching a checklist.

Strong candidates will have experience in several of:

• machine learning and statistical modeling;
• causal inference and experimentation;
• recommendation, advertising, pricing, marketplace, credit, or other decision systems;
• bandits, reinforcement learning, optimization, or active learning;
• uncertainty estimation;
• counterfactual evaluation;
• production ML systems;
• Python, SQL, and large behavioral datasets.

Why This Role Is Different

Most ML systems learn from a dataset.

Here, **the decisions made by the model influence the data the model sees next**.

That creates a continuous loop:

**Decision → intervention → outcome → learning → better decision**

The long-term opportunity is to build that capability once and apply it across a portfolio of businesses and increasingly broad commercial decisions.