Staff Software Engineer, AI Platform
About The TeamWe are building JLL's internal AI platform: the layer every team in the firm uses to ship AI agents and AI-backed products. That means a gateway that puts a large and growing set of models from multiple providers behind one API, a shared chat surface that makes any agent reachable by employees across the firm, the developer tooling that takes an agent from an idea to something running in production, etc.You would join early enough to shape what this platform becomes. The foundations are live and already carrying production traffic, and the decisions still ahead of us are the ones that determine how every team at JLL builds with AI. Few engineers get the chance to set that direction at a firm this size, and fewer still get to see it in use across the business within weeks of building it.About The RoleThis is a Staff level individual contributor role on a young platform. You will set technical direction rather than receive it: choosing the abstractions other teams will build against, deciding what belongs in the platform and what does not, and being accountable for whether those choices still look right a year from now.The work is genuinely both halves of the title. It is serious distributed systems engineering, with a gateway on the critical path of everything the firm builds, and it is AI engineering, where the hard problems are model routing, evaluation, guardrails, token cost and latency. We need someone who has shipped LLM backed systems in production and who would also be a strong platform engineer on any team.What You'll DoBuild the platformOwn significant parts of the platform end to end, from the shape of the API through to how it behaves under loadBe accountable for reliability and performance in production, on a gateway that sits on the critical path of everything the firm builds with AIDesign the model gateway: provider abstraction, routing, failover, and making a provider switch a configuration change rather than a migrationMake the agent surface work regardless of which framework a team chose or where their agent runsSolve the AI engineering problemsBuild the evaluation and benchmarking capability the platform needs, largely from scratchDesign the guardrails that let an agent reach production safely, and help turn draft internal standards into something enforceable in codeOwn inference cost and latency as first class engineering concerns, including spend attribution and the routing decisions behind itKeep pace with a model landscape that changes weekly, and make onboarding a new model routine rather than a projectSet the standardMake the compliant path the fast path, so that building on the platform is how a team clears architecture, security and responsible-AI reviewBe the engineer other teams bring their hardest integration problems toRaise the bar on observability before scale forces the issueMentor across a distributed team, and leave behind designs and documentation that outlast your involvementRequiredWhat we're looking for8+ years building and operating production systemsDirect experience shipping agentic systems to production, not prototypes: agentic patterns, retrieval, tool calling, inference cost and latency, etc. Hands-on experience with more than one model provider, including dealing with the differences between them in practiceExperience with evaluation, observability or guardrail tooling for agentic systemsStrong platform engineering track record, on systems where other engineering teams were your usersHands-on depth in at least one major public cloud, including running and debugging production workloadsExperience designing APIs and abstractions that others build against and that you then have to keep stableDemonstrable production ownership: on-call, incident response, root cause analysisA track record of pushing work through a large organisation: navigating process, winning the argument, and getting decisions unblocked rather than waiting on themExcellent English communication, written and spokenPreferredBuilt or operated an internal developer platform or shared multi-tenant serviceKubernetes and container-based deliveryInfrastructure as codeEnterprise security, data privacy or responsible-AI review processesHow we work: A small team that moves quickly inside a very large organisation, where the fastest route is rarely the obvious oneAn early stage platform where the roadmap is still being writtenDeciding without complete information, and revisiting the decision when it turns out to be wrongA distributed team with limited timezone overlapProduction users from day one, on a platform still being built underneath them