Lead Data Scientist

The world is in a waste crisis. Currently we produce 2.1 billion tons of solid waste per year. Data collection of the waste we produce is non-existent, meaning no systematic transparency and no accountability. It means that recycling targets are not upheld, dumping of waste into our oceans remains nobody’s responsibility, recyclables get sent to landfill or incineration, and producers have little visibility of how their packaging performs at end of life. Recycling rates remain at 10% and, unless we change, by 2040 the plastic stock in the ocean will have quadrupled - a problem that already costs society $1.5 trillion each year. Our mission is to increase transparency and automation in waste management to accelerate the circular economy. Greyparrot's camera and AI systems generate granular, real-time waste composition data across facilities worldwide. That data is only valuable if it can be turned into insight that clients trust and act on.

We are looking for a Lead Data Scientist to act as the “engine room” of our data methodology, transforming raw computer vision outputs into insights the industry can rely on. The near-term priority is delivery; owning the contracted Deepnest client analytics, the reports and data clients have paid for, shipped on time and to a high standard. The metric engine work is what makes delivery scalable and trusted; developing the statistical methodology that turns raw computer vision outputs into defensible, quantified waste metrics. Both matter from day one. Longer term will support other customer & marketing facing work as well. You will report directly to the CTO and have one direct report, Data Analyst, with scope to grow as the business scales. You will sit alongside the Head of Data R&D who owns longer-term structural research upstream of the metric engine.

The data is physical-world data: noisy, incomplete, and derived from computer vision models running in live recycling facilities. This is not a clean-warehouse role. It suits someone who treats messy data as the problem to solve, not a reason to wait.

Outcomes:

• The metric engine advances: The next iteration of Greyparrot's statistical modelling framework is implemented; strengthening how we reconcile process flows, extrapolate across coverage gaps, and quantify confidence in outputs. At 12 months, a credible path toward a confidence-aware, probabilistic foundation is underway. The methodology is documented, defensible, and ready to be productionised by ML Ops.
• Contracted deliverables ship on time, every time: Deepnest clients receive their analytical reports and insight outputs to a consistently high standard and on schedule. The methodology behind the numbers is defensible, the findings are actionable, and clients trust what they receive. There are no surprises at delivery.
• Insight delivery is repeatable, not heroic: A documented framework – templates, quality standards, methodology – exists so output quality does not depend on starting from scratch each engagement. The process is written down, transferable, and does not live in your head.
• R&D and delivery are in sync: The Head of Data, R&D has a clear, consistent picture of which model outputs translate to client value. You provide that feedback loop reliably, and it shapes what gets prioritised on the research roadmap. There is no gap between what the models produce and what clients actually need.

Experience & Background:

• Physical-world data: 5+ years in data science working with large-scale, noisy real-world data – environments where data quality and fail modes are constant challenges.
• Background in high-volume, complex real-world data industries: satellite and geospatial, weather forecasting, industrial IoT, manufacturing, is a bonus.
• Deep learning familiarity: A strong practical understanding of the implications of working with data derived from deep learning models; specifically the nuances of integrating computer vision outputs into broader statistical simulations. You know where the model can mislead you and how to account for it.
• Python and SQL: You build analysis pipelines and get to robust outputs independently, without needing a data engineering team to do it for you. This is not an expectation of production ready output.
• Owned external deliverables: Reports or data products that clients or senior stakeholders have relied on. You understand what makes insight land versus what gets ignored.
• Built from scratch: You have built methodology and process where none existed, not just inherited and executed. You are comfortable setting standards and navigating ambiguity at pace.
• People leadership: You have managed or mentored at least one person and have a clear view of what good looks like. You can set a standard and give others the structure to work within it.

What Success Looks Like:

• 90 days: You have owned at least one contracted Deepnest deliverable end-to-end. Your team has clear scope and is working effectively. You have a clear picture of the current methodology, where the metric engine stands, and where the gaps are.
• 6 months: The next iteration of the proprietary metric engine and data modelling framework is implemented. A repeatable delivery process is in place and documented.