Staff Data Scientist - Network

Relay is fundamentally reshaping how goods move in an online era. Backed by Europe's largest‑ever logistics Series A ($35M), Relay is scaling faster than 99.98% of venture‑backed startups. We're assembling the most talent‑dense team the logistics industry has ever seen. Relay's Mission is to free commerce from friction. Today, high delivery costs act as a hidden tax on e-commerce, quietly shaping what can be sold online and limiting who can participate. We envision a world where more goods move more freely between more people, making the online shopping experience seamless and accessible to everyone.

Team: ~110 people, more than half in engineering, product and data; 45+ advanced degrees across computer science, mathematics and operations research; Thousands of data points captured, calculated, analysed and predicted for every single parcel we handle. An intellectually vibrant culture of first-principles thinking, tight feedback loops and relentless experimentation.

About the role: Relay’s network runs on forecasts. Every shift released in sortation, every middle‑mile van dispatched, every last‑mile route planned, every expansion decision made - all downstream of models that predict how parcels move through our system. When those models are right, the network runs efficiently and cost per parcel drops. When they drift, the cost compounds across every stage of the operation. The Network squad builds and maintains the forecasting engine that powers all of it, and Demand Forecasting is its core: one forecast of what will be available to sort, at outcode granularity, from D0 out to D30.

As a Staff Data Scientist, you are the technical anchor for Demand Forecasting. You own the hardest and most ambiguous parts of the forecast, and you set the methodology and validation standards the rest of the area works to. That means owning the single integrated forecast of what volume the network will have to move, by area and out to thirty days, which the demand‑management layer then turns into the operational plan the sort centres and transport teams run on. It means owning the model‑driven end of that forecast, where the horizon runs past any live tracking data and expected parcels have to be generated from models rather than observed. It means owning the forecast of inbound international volume, one of the hardest prediction problems we have. And it means owning the models that predict each parcel's size and weight, which turn a parcel count into the physical volume that actually has to be sorted and loaded.

This is a hands‑on role. You will spend most of your time building, not managing. You set the direction for how Demand Forecasting models, evaluates and ships its work, and you raise the technical bar across the area, but you do it as the most senior individual contributor in the room, on the tools. You'll work alongside a Senior Data Scientist who owns the domestic volume forecasts, an ML Engineer who keeps the models running reliably in production, and an Analyst who owns forecast accuracy and data quality. People leadership, strategy and cross‑squad priorities sit with the Data Science Manager who leads the squad; you own the science.

Relay operates a centralised data team of around 30 Data Engineers, Analysts, and Data Scientists, with specialists embedded into squads across the business. You will sit in the Network squad and report into the centralised data team. Demand Forecasting is growing, and as its Staff DS you will have the deciding voice on its technical direction and the modelling approaches it adopts.

What You’ll Do:

• Own the integrated forecast end to end. Blend live tracking signals in the near term with model-generated parcels further out into a single view, by area, from today to thirty days ahead. This is the view the demand‑management layer turns into the plan that Sortation, Middle Mile and Last Mile actually run on.
• Build the hardest models in the area. The model‑generated long‑horizon forecast, the inbound‑international volume forecast, and the parcel size and weight models that make the forecast a measure of physical volume rather than just a count. These are the ambiguous, high‑leverage problems, and they are yours.
• Set the methodology and validation standards for Demand Forecasting. Define how models are evaluated, how accuracy is measured at each horizon (a forecast made thirty days out shouldn't be held to the standard of one made two days out), and what 'good' looks like across the area's models.
• Raise the technical bar. Review approaches, make the model-choice and build‑vs‑buy calls, and mentor the Senior Data Scientist and Analyst alongside you, without taking on their line management.
• Define the forecast's interfaces. Decide what Demand Forecasting hands to the demand‑management layer, to Routing, and to the network‑planning function: at what granularity, with what guarantees, and how error is attributed when a number turns out wrong.
• Learn the operational processes your models serve, supported by the squad and the teams who use the forecasts, and identify where the current approach falls short.
• Own production quality across the estate, working with the ML Engineer so that models are monitored, drift is caught early, and accuracy problems are traced to the right cause rather than re‑tuned blindly.
• Work with Finance, who extend the operational forecasts into longer‑range financial projections, to keep the handoff between operational and financial models reliable.
• Quantify the impact of model error on cost per parcel, and use it to decide where the area invests effort.

Who Will Thrive in This Role?

• You have been the technical anchor on a modelling team before. You've owned the hardest problems, set the standards others worked to, and been the person the team turned to when an approach needed a call. You did it hands‑on, as a senior IC, not by moving into management, and that's the path you want to keep on.
• You think in interconnected systems. A demand forecast isn't just a number; it drives how many shifts are opened, how many vans are dispatched and how routes are planned, and it feeds a downstream planning decision and the network's expansion models. You care about how your models connect to the models around them, and you design their interfaces deliberately.
• A deep track record of building and delivering models from ambiguous starting points. You understand the problem, build something useful, validate it against real operations, and iterate. You evaluate models well beyond standard offline metrics, connecting outputs to downstream applications and business KPIs and measuring how improvements translate into operational impact.
• Strong Python and SQL, and depth across the full modelling lifecycle - from data extraction and feature engineering through training, validation and production deployment. You've worked with time‑series forecasting across classical statistical approaches, gradient boosting and deep learning, and you understand the trade-offs well enough to make and defend the choice for a given problem. You'll be supported by a dedicated ML Engineer, but you set the standard for how the area's models are built.
• At least 8 years in a data science or quantitative modelling role, with clear examples of models you built that informed operational or commercial decisions, and of methodology or technical direction you set for a team. You know a model isn't done when it trains well; it's done when it's running, monitored, and trusted.
• You communicate with non-technical stakeholders with authority. The squads that consume the forecasts need to trust them, and that trust comes from explaining what the models do, where they're reliable, and where they're not, without hiding behind the maths.
• You're comfortable using AI tools - LLMs, code assistants, and similar - to accelerate your workflow, from exploratory analysis to code generation, and you're curious about where these tools can augment the modelling process itself.
• This role suits someone who wants to see whether their models made a real difference to how the network operates. There is a direct feedback loop between your work and operational outcomes, and as the Staff DS you own the most consequential end of it.
• Logistics or delivery network experience is a plus, but what matters more is the ability to learn a complex operational domain quickly and model it well.

Compensation, Benefits & Workplace:

• Generous equity, richer than 99% of European startups, with annual top‑ups to share Relay's success.
• Private health & dental coverage, so comprehensive you'd need to be a partner at a Magic Circle law firm to match it.
• 25 days of holidays.
• Enhanced parental leave.
• Located in Shoreditch, our office set-up enables the kind of in‑person interactions that drive impact. We work 4 days on‑site, with 1 day remote.
• Hardware of your choice.
• Extensive perks (gym subsidies, cycle‑to‑work, Friday office lunch, covered Uber home and dinner for late nights, and more).

Who Thrives at Relay?

• Aim with Precision: You define problems clearly and measure your impact meticulously.
• Play to Win: You chase bold bets, tackle the hard stuff, and view constraints as fuel, not friction.
• 1% Better Every Day: You believe that small, consistent improvements lead to exponential growth. You move quickly, deliver results, and learn from every experience.
• All In, All the Time: You show up and step up. You take ownership from start to finish and do what it takes to deliver when it counts.
• People-Powered Greatness: You invest in your teammates. You give and receive feedback with care and candour. You build trust through high standards and shared success.
• Grow the Whole Pie: You seek out win‑win solutions for merchants, couriers, and our customers, because when they thrive, so do we.

Relay is an equal‑opportunity employer committed to diversity, inclusion, and fostering a workplace where everyone thrives.