Forward Deployed Engineer, Business/Domain

Atmospheric AI Limited; Greater London; Hybrid; GBP 90,000 - 140,000; Full time

Job summary

Atmospheric AI Limited is seeking a practical, senior-leaning problem-solver who can extract buying logic from complex data and encode it into our platform. You will capture approved suppliers, substitutions, decision rights and risk policies, turning them into a defensible model grounded in real data.

You will lead discovery with executives, build a simple cost view, and write a clear value case with named assumptions.

Qualifications

• Proven ability to extract buying logic and build the model yourself.
• Experience structuring ambiguous problems and presenting numbers to CxOs.
• Category strategy with quantified savings at named clients.

Responsibilities

• Capture approved suppliers, substitutions, decision rights, switching costs and risk policies and encode them in the customer's model.
• Perform category analysis on client data: cost build-ups, regional comparisons, carbon and exposure by route.
• Develop value cases in P&L terms with named assumptions and defensible baselines.
• Strategic calls: supplier-base redesign, decarbonisation pathways, make-versus-buy, quantified assumptions.
• Discovery, bootcamps and reviews: hypothesis-led discovery, live-data build sessions, reviews that close the loop from recommendation to action.
• Expansion and the playbook: identify next value pools and document learnings for faster deployments.

Skills

Buying logic modeling

Category analysis

Strategic decision making

Job description

Everything physical is made from raw materials: polymers, chemicals, metals and minerals, a $16T economy that software has barely modelled. Atmospheric AI computes what those materials should cost, what carbon they carry and where their supply is exposed, from the physics of how they're made, so manufacturers can decide and negotiate from ground truth.

The problem

A category manager running nine-figure spend knows which suppliers are approved in practice, which substitutions engineering will sign off and where the switching costs sit. Almost none of it is written down. Our platform can compute what a material should cost, the carbon it carries and where supply is exposed, but it needs that buying logic to answer the customer's own question. You get it out of their heads, codify it into the platform yourself, and turn it into decisions their finance team will sign.

What you'll own

• Buying logic. Capture approved suppliers, substitutions, decision rights, switching costs and risk policies, and encode them in the customer's model with your own hands. Their experts correcting the edge cases is how you know you've got it.
• Category analysis on their data. Cost build-ups, comparability across grades and regions, carbon and exposure by route, on the customer's own data with its supplier-name mess and mixed units.
• Value cases in P&L terms. Named assumptions and a defensible baseline, carried through the conversations where the commercial outcome is decided.
• Strategic calls. Supplier-base redesign, dual-sourcing, decarbonisation pathways, make-versus-buy, quantified assumption by assumption.
• Discovery, bootcamps and reviews. Hypothesis-led discovery, hands-on build sessions on live data, and reviews that close the loop from recommendation to action.
• Expansion and the playbook. Find the next value pool in each account, and write down what you learn so the next deployment runs faster.

What we're looking for

• You've extracted buying logic and built the model yourself, professionally. Both halves. Workshop-and-handoff doesn't clear it, and neither does building to someone else's spec. We can teach our platform; we can't teach that.
• Top-tier strategy or operations consulting craft, or its equivalent. You've structured ambiguous problems, built the fact base yourself and put a number in front of a CxO that survived being pulled apart.
• Category strategy with receipts. Should-cost build-ups, category strategies and negotiation cases with quantified savings at named clients, and you can say which assumption was load-bearing.
• Presence in the room. You've run discovery with senior stakeholders and held your ground when the numbers were challenged.
• AI-native. LLM tooling is daily leverage, and codifying expert judgement into something an agent can run is the part you enjoy.

Production code isn't required; engineers have your back there. Three lines we hold: there are no SOWs or utilisation targets, your output is decision-grade work tied to named value, and you build on the platform while its core stays with the product team.

The take-home: from a small category dataset, extract the buying logic, build a simple cost view, make one strategic call and write the value case with named assumptions.