Lead Data Scientist - Freight Intelligence

Lead Data Scientist \- Freight Intelligence Remote Portfolio roles Remote Full\-time **About Overroute:** Overroute builds AI agents for freight operations. Our agents work inside the tools operators already use, such as email and chat, to track shipments, answer operational questions, communicate with drivers, set appointments, and draft customer communications for freight carriers. **The role:** As Lead data scientist, you set the technical direction for data science at Overroute. Youʼll build the tools our agents call, choose the methods, and build, deploy, and scale the models and tools behind it in production for large fleet operators. You will work with deep, real\-world operating data from day one, alongside the operators who use the product, and own your work end to end, from first baseline through production monitoring. This is a senior individual contributor role. You lead through technical ownership and hands\-on delivery rather than by managing a team, and you decide when a model is ready to ship. You work agent\-first: AI coding and data agents are part of how you do data science every day, so one person can cover the ground of a team. **What you'll own:** * Reliable datasets and labels. Define ground truth from operational sources that may disagree or be incomplete, and build datasets for evaluation and modeling. * Predictive modeling and decision support. Work with Product, Engineering,and operators to turn freight problems into models and tools that support operational decisions. * Production models at scale. Take models from the first baseline to an initial production release and scale them in live workflows, partnering with Engineering on evals, deployment and infrastructure. Own automated monitoring, retraining, validation, and model updates, balancing accuracy, response time, reliability, and cost. **How we work:** * Simplest method first. Compare quality, response time, and cost against simpler alternatives. A model earns complexity by beating a published baseline. * Ground truth before models. Define the label before training or grading anything. * Language models do not do the math. Calculations come from deterministic code. * One platform, many customers. Customer differences are configuration, not code branches. * Clear ownership. Engineering owns data pipelines, infrastructure, and security implementation. You own labels, methods, models, and evaluation. Product owns priorities, customer policy, and approval requirements. * Make uncertainty clear. Work with Product and Engineering to make each prediction's confidence, limitations, and missing inputs clear to operators. * Measure customer impact with Product. Connect model accuracy to operational outcomes and customer return on investment ROI. Compare outcomes against a baseline and account for the cost of running and maintaining the models. **What you bring:** * 5\+ years of hands\-on experience in data science or applied machine lear