AI Platform Engineer

About Saturation

Intelligence works best when it understands how things are connected: how a creative decision changes the plan, how the plan changes the money, and how both shape what happens next.

That is the thesis behind Saturation. We are building the intelligent operating system that will power the next generation of financial workflows. AI moves beyond the chatbot and into the work itself, helping productions anticipate what is coming, make better decisions, and work in ways that were never thought possible.

More than 17,000 producers have budgeted over $2.7 billion with us, from independent productions to projects made with some of the world’s largest studios and brands. Now we are bringing the next generation of Saturation to web, mobile, desktop, and agents.

About the role

We are looking for an AI Platform Engineer to build the intelligence platform behind Saturation. The work sits where product judgment meets systems engineering: deciding what context matters, how it should be represented, which tools models can use, and how the result becomes dependable product behavior.

The hardest problems are larger than a prompt. They reach into compilers that turn messy production documents into useful structure, memory that preserves relationships across documents and live state, state machines that keep long-running work recoverable, and evaluations that tell us whether the system is actually getting better.

You will help decide which capabilities are worth building, prove them in real product use, and stay close enough to know when they work.

What you'll do

• Build the context layer that connects documents, live product state, conversation, and the relationships the system needs to remember.
• Design tools and execution paths that turn model decisions into reliable, permissioned product behavior.
• Design benchmarks and evaluations that measure useful behavior, expose failure modes, and guide what we build next.
• Build long-running agent workflows that remain recoverable, steerable, and understandable.
• Define where people review, correct, or take control without breaking the flow of the work.
• Trace failures across models, tools, state, and product behavior until you find the real cause.
• Ship into the product and learn from how people use what you made.

How we build

We are a fully distributed international team with no central office. We work asynchronously so there is room for focus. We write decisions down, share work before it is polished, and speak directly when something does not feel right.

AI is not isolated in a research group. The people building it work across models, infrastructure, domain systems, and the experience people touch.

What we're looking for

• You have owned a production language-model system through failures, iteration, and changing model behavior, not only built prototypes.
• You have strong instincts for what context a system needs, when it needs it, and what should remain outside the model.
• You care about tool design and can reason clearly about the boundary between models and software.
• You turn ambiguous behavior into benchmarks, traces, and evaluations that a team can use to make decisions.
• You can move from a deep systems problem to the product experience it makes possible.
• You explain technical tradeoffs clearly and can make progress when the path is not yet defined.

Our tech

Models are one part of the system. Around them, we build structured context, typed tools, compilers, relationship-aware memory, durable execution, and evaluations we can run again and again. The work is deciding what belongs in software, what belongs in the model, and how both can improve without making the product feel more complicated.

Helpful experience

Experience with any of the following may help:

• Agent tool-use systems or evaluation infrastructure
• Document compilers or intermediate representations
• Relationship-aware memory or knowledge graphs
• State machines or durable execution
• High-trust product systems

The offer

We built Saturation as a remote company from the beginning. There is no headquarters to orbit and no default city. This is a full-time role open internationally, so you can work from wherever you do your best work.