Staff AI Product Builder

Atmospheric AI Limited; Greater London; Hybrid; GBP 110,000 - 180,000; Full time

Benefits offered by this job

Competitive compensation

Job summary

Atmospheric AI Limited in London seeks a senior software architect to design and build the core system powering our physics-based materials insights platform. You will own architecture, data contracts, and the agent layer, shaping production services from the model to the product frontend.

You will lead end-to-end features, implement robust tests, observability, and reliability, and collaborate with product teams to outline what to build next.

Qualifications

• Experience designing production-grade systems.
• Ability to own and communicate system architecture decisions.
• Strong back-end skills with Python and AWS.
• Familiarity with TypeScript/React for frontend touchpoints.
• Experience with AI/LLM-enabled workflows.

Responsibilities

• Design the system architecture and data contracts between services, agents, and product.
• Own the agent layer, planning work and ensuring correctness before delivery.
• Build end-to-end product features, including front-end work where needed.
• Ensure tests, observability, performance, and reliability at scale.
• Mentor peers and raise engineering standards across the team.

Skills

System architecture

Python

TypeScript

React

AI/LLM systems

Production experience

Backend development

Tools

Python

FastAPI

Postgres

Dagster

AWS

LangGraph

Pydantic AI

Next.js

TypeScript

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

Category managers run nine-figure spend on supplier quotes and benchmarks they can't see inside. Cortex works alongside them: agents gather the data, run it through our physics model and prepare the analysis, so the buyer's time goes on judgement and the negotiation. You'll design that system and build most of it yourself, from the services underneath to the product people use every day.

What you'll own

• System architecture. The services, data contracts and APIs between our physics model, the agents and the product. You set the patterns the rest of the team builds on.
• The agent layer. How work gets planned, handed to agents and checked before it reaches a person, with models as swappable parts and the logic in code we control.
• Product, end to end. Features from first sketch to production, including the front-end work needed to land them. You talk to users and decide what to build next.
• Correctness you can measure. Tests and evals that catch a wrong answer before a customer does, and observability that shows cost, latency and failures.
• Reliability as we grow. Systems that stay fast and predictable as customers, data and agents multiply.
• The engineering bar. Code review, standards and tooling that make everyone around you faster.

What we're looking for

• A strong software architect who still builds. You've designed systems that ran in production for real users, and you wrote much of the code yourself. Walk us through one and why it held up. That's where every conversation with us starts.
• Product sense. You've built software people chose to use, and you can name a decision you made because of how they behaved.
• Mostly backend, comfortable anywhere. Expert Python or a comparable language, and resourceful enough in TypeScript and React to finish a feature.
• You want to build with AI. Production agent experience helps. If yours is prototypes and side projects, show us what you built and what you learned.
• You raise the bar around you. You've set standards, mentored, and made a team faster through the systems you built.

Nice signals: production LLM or agent systems; evals; data-intensive or scientific software; data science; supply-chain, commodities or manufacturing exposure.

Stack: Python, FastAPI, Postgres, Dagster, AWS; LangGraph, MCP, Pydantic AI; Next.js and TypeScript. A signal, never a requirement.

The take-home: a short task shaped like the real work, and later a working session on a problem like ours.