Senior Domain Engineer, Trading

Our mission is to make the raw material world computable.

Raw materials — the polymers, chemicals, metals and minerals that become every product on earth — power a $16T+ physical economy, but they remain largely unmodeled and uncomputed in software. Decisions that move billions are still made on averages, assumptions and opaque supplier narratives.

About the role

We're hiring into our trading pod — the part of the platform that gives procurement, sustainability and supply-chain teams defensible answers to questions like what will my polymer cost next quarter, how does a Brent shock cascade through my supply chain, or what happens if the Strait of Hormuz closes. Your job is to make those answers true.

You'll do that by building the quantitative models, datasets and services that turn messy commodity-market reality — Brent prints, Argus assessments, refinery outages, ethane economics, regional benchmarks — into structured signals the platform reasons over and the AI agent calls as tools. Not a pure data scientist who consumes datasets, and not a pure backend engineer who plumbs them: someone who builds the dataset, builds the model, and ships it as a service.

We're looking for someone who comes with experience modeling these systems. You've done this work professionally — on a trading desk, at a fund, at a price-reporting agency, or in a petrochemicals consultancy. You already know why ethylene cascades from Brent in Northwest Europe but decouples at Mont Belvieu, why ethane versus naphtha cracker economics diverge, and how an Argus or Platts assessment is actually constructed — because you've had to be right about it for people with money on the answer. We can teach our stack; we can't teach years in these markets. The role will broaden well beyond petrochemicals — this is where it starts.

What you'll do

• Model commodity price cascades — codify how Brent moves NWE naphtha moves European ethylene; ethane decoupling at administered-ethane sites; regional benchmark assignment
• Forecast and stress-test under disruption — time-series models that project prices as distributions, not points, and simulation to study how shocks propagate through the supply-chain graph
• Construct pricing datasets with provenance and uncertainty — sources, confidence and lineage tracked first-class. You own your datasets end-to-end, including knowing what's missing.
• Ship scenario and sensitivity engines — "what happens to my polymer cost if crude hits $X?" should return a structured answer with explicit assumptions and confidence bands, from a production service the agent can call
• Build for repeatability — you're not just shipping models, you're shipping the system that ships models
• Close the loop with customers — the deliverable isn't a model, it's a customer getting a defensible answer they act on
• Be in the room — expect to walk a procurement lead or chemical specialist through your methodology and where confidence is high versus low. We don't ship black boxes.