Senior AI Platform Engineer - Knowledge & Reasoning

Senior AI Platform Engineer -  
Knowledge & Reasoning 

Location: Flexible - remote, hybrid, or based at an Esko location.  
Worldwide locations are considered.  
Function: Software Engineering 
Reports to: Team Leader, Esko AI 

 
About the role 

The Esko AI team is a group of specialists focused on leveraging AI to solve valuable and often challenging customer problems that have historically resisted automation via purely rule-based approaches. To succeed, we must ingest and leverage expert knowledge to augment the capabilities of our AI agents and services. This capability will be a key part of Esko’s centralized AI platform, enabling wider teams to build AI into their products using a common base. 
 
Packaging workflows are governed by knowledge of very different kinds: regulatory requirements (FDA, EMA, QRD), technical specifications, brand guidelines, customer-specific conventions, and practices that currently live in the heads of experts. Some of this knowledge is binding, some advisory, some contradictory, and much of it applies only to a particular customer, brand, or specific range of designs. 

For AI systems to operate reliably in this environment, they must do more than retrieve relevant information. They need to determine which knowledge applies, distinguish mandatory requirements from guidance, handle missing or conflicting evidence, and recognize when human judgement is required. 

We are building knowledge and reasoning as a shared capability of Esko's AI Platform, enabling product teams to create AI-powered workflows grounded in trusted, contextual knowledge. As part of the platform team, you will help shape and build this capability, taking end-to-end ownership of substantial technical areas. 

What you will do 

• Design the knowledge representation: how constraints, transformations, procedures, and guidelines are modelled so that both agents and deterministic validators can use them, and so that humans can read, curate and correct them. 

• Build and operate platform services for ingesting, curating, retrieving, and versioning knowledge, ensuring it is scoped correctly to the relevant customer and workflow context. 

• Define how semantic retrieval, deterministic validation, and human review work together, including how applicability, uncertainty, conflicting evidence, and escalation are handled. 

• Design and implement reusable APIs and tools that allow AI systems and deterministic services to apply, validate, and explain contextual knowledge. 

• Build evaluation suites covering retrieval quality, applicability, conflict and ambiguity handling, and explanation quality, using measurable results as the primary evidence of progress. 

• Take end-to-end ownership of substantial technical areas, including their security, tenant isolation, reliability, observability, performance, and compatibility for consuming teams. 

• Work directly with domain experts and product engineering teams to turn informal expertise into structured, testable knowledge with clear provenance, citation, and release practices. 

What we are looking for 

• Strong software engineering experience designing, building, and operating production systems in Python, TypeScript, Java, or a similar language.  

• At least 5 years’ experience in a software engineering role, preferably including ownership of a platform, API or other shared-services product used across multiple internal teams.  

• Experience building knowledge, retrieval, or rule systems that other teams depend on, and the judgement to choose the right formalism for a problem, whether that is a schema, decision table, rules engine, graph, or retrieval. 

• Practical experience with RAG, information extraction, and structured validation (JSON Schema, Pydantic, or equivalent). 

• Sound judgement when reasoning about applicability, exceptions, missing evidence, and ambiguity, including recognizing when a system should decline to answer or require human review. 

• Experience constructing evaluation datasets, and the ability to write clearly about modelling decisions, tradeoffs, and failure modes. 

• Undergraduate, Master’s degree or PhD in Computer Science / Machine Learning / Data Science / Artificial Intelligence, or related disciplines. 

• Comfort operating at the boundary between a platform team and the domain engineering teams that consume it, balancing shared needs and central objectives against individual team priorities and timelines. 

• Excellent written and verbal communication, with the ability to work fluently with both technical and business stakeholders alike. 

Preferred experience 

• Semantic and symbolic tooling (RDF/OWL, SHACL, Datalog, DMN, production rule engines, policy-as-code).  

• Agentic systems with tools calling and validation-before-action.  

• Multi-tenant SaaS with strict isolation and audit requirements.  

• Experience in a high-accountability domain such as pharmaceutical or food labelling, and legal technology. 

• Experience in packaging graphics or the wider graphic arts industry. 

What success looks like 

• Delivery of the knowledge module, conforming to agreed specifications and architecture, in a way that delivers measurable improvements in AI agent and service effectiveness (e.g. increased end-user productivity).  

• Product teams build shared knowledge services rather than constructing their own.  

• Agents apply the correct tenant-, market-, and customer-specific constraints with measurably high recall, can explain which knowledge items influenced a recommendation or validation result, and escalate rather than guess when knowledge is ambiguous, conflicting, or absent.  

• Tenant-specific knowledge never crosses tenant boundaries.  

• The knowledge base itself is versioned, tested, legible – and curatable by the humans who must ultimately stand behind it. 

• Delivery of initiatives occurs in line with Esko’s Software Development Lifecycle and in accordance with company policy and security best practices.  

• The knowledge module grows as a coherent whole rather than fragmenting into one-off, team-specific solutions.