Senior AI Engineer - Exploration & Prototyping

we are looking for a Senior AI Engineer - Exploration & Prototyping.
That is this role. You take an open question, run a focused spike, and come back with numbers and a recommendation. You read the source of the frameworks you evaluate rather than trusting their marketing. You build prototypes to settle arguments.
You do not own a subsystem and you do not ship to customers, which is exactly what protects the work: exploration inside a delivery team always loses to the sprint. You sit alongside the platform, research, and forward-deployed teams, you borrow their context freely, and your output is evidence they can act on.
You will be trusted with real influence early. The recommendations you write become the architecture other people build against, so the bar is not a working demo but a defensible conclusion, including the ones that say no.
What Youll Do:
Run technical spikes that close open decisions, covering agent orchestration frameworks, real-time transport, memory protocols, agent interoperability standards, LLM selection and routing, evaluation harnesses, and the production library and stack choices underneath all of it.
Build prototypes to de-risk, standing up something real quickly, proving or disproving the thing in question, and moving on without becoming attached to the code.
Read and evaluate unfamiliar codebases, going into the source of a candidate framework to find out whether it can actually support what we need rather than what its documentation implies.
Design the measurements that make a decision defensible, building the harness, running the comparison, and reporting latency, cost, and failure behavior honestly.
Own build-versus-adopt recommendations for platform infrastructure, frameworks, and libraries, including a clear statement of what it would cost to be wrong.
Write the recommendation down. Every spike ends in a short, decisive document another engineer can act on, with the evidence, the rejected options, and the reasoning behind the call
Hand off cleanly, transferring what you learned to the team that will own the capability in production, and staying available while they pick it up.
Track the landscape across agentic infrastructure, real-time frameworks, and adjacent AI tooling, and bring forward the things that genuinely change what we can build.