> Markdown version of [/jobs/ext/4597463-principal-embedded-engineer-ai-solutions](https://www.wearedevelopers.com/jobs/ext/4597463-principal-embedded-engineer-ai-solutions). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md). --- # Principal Embedded Engineer, AI Solutions - **Company:** Autodesk - **Location:** Barcelona, Spain - **Contract:** Permanent contract - **Skills:** LangGraph Framework, Training Data, Java (Programming Language), Application Programming Interfaces (APIs), Artificial Intelligence, Amazon Web Services, Automation of Tests, Microsoft Azure, C Sharp (Programming Language), Cloud Computing, Code Generation, Software Quality, Code Reuse, Continuous Integration, Customer Data Management, Data Stores, Query Languages, Graph Database, Python (Programming Language), Neo4j, Software Product Management, Rapid Prototyping Process, Software Deployment, SPARQL, Systems Integration, TypeScript, Management of Software Versions, Enterprise Data Management, Enterprise Software Applications, Multi-Agent Systems, Model Validation, AgentCore, AI Coding Agents, Backend, Langfuse — LLM Observability and Analytics Platform, Agentic-AI, Cloudformation, Data Layers, Build Management, Data Lakes, LangSmith, Production Code, RAGAS (Retrieval Augmented Generation Assessment), Bicep, Cosmos DB, Entity Resolution, Graph RAG, Software Coding, Terraform, Autodesk Autocad, Data Pipelines, Serverless Computing, Human in the Loop - **Published:** November 6, 2026 - **Apply:** https://dejobs.org/x/x/BFA30CD3AD564D94845AAAAE14BE94C6/job/ ## About the Role The successful candidate combines strong software and AI engineering skills with curiosity, customer empathy, business judgment, and the ability to operate independently when the problem is not yet well defined. This role is engagement-driven with travel up to 75% during active engagements, including extended periods working onsite with customers., * 6+ years building and operating production software, including 2+ in a customer-facing or embedded role (forward-deployed or founding engineer). CS/Engineering degree or equivalent hands-on experience. * Full-stack range. Expert Python plus one of TypeScript, Java, or C# - comfortable across backend, data pipelines, integrations, and a lightweight React surface when a workflow needs one. * Cloud fluency in Azure or AWS: containers, serverless, managed data stores, and identity and networking inside customer-governed environments (e.g., infrastructure as code with Terraform, Bicep, or CloudFormation), with strong API, distributed-systems, CI/CD, and observability fundamentals. * Agentic systems in production. You've built and run single- and multi-agent systems - tool schemas, memory, human-in-the-loop checkpoints - on a modern orchestration framework (e.g., LangGraph, LangChain, Bedrock AgentCore, etc), MCP making practical cost and latency trade-offs (e.g., model selection, caching). * Evaluation as engineering. You build golden datasets and eval harnesses from real customer data, and you measure groundedness, task success, and retrieval quality using standard tooling (e.g., RAGAS, LangSmith, Langfuse). * Retrieval and data readiness. You've built retrieval systems end-to-end (chunking, embeddings, hybrid search, reranking) over a vector store and can assess and fix messy enterprise data - lineage, permissions, entity resolution, schema mapping. * Knowledge architecture for AI. You model domain semantics into AI-ready context - ontologies, semantic layers, knowledge graphs - and ground agents in it via GraphRAG or equivalent, using at least one graph technology (e.g., Azure Cosmos DB Gremlin, Amazon Neptune, Neo4j, or equivalent) and its query language (e.g. Gremlin, Cypher, SPARQL). * Ownership under ambiguity. You take greenfield customer problems from framing to production and measured outcomes and can explain a non-deterministic system with equal credibility to an engineer and an executive., * Domain fluency in AECO, with the ability to ramp quickly on design-to-construction-to-operations processes. * Autodesk platform fluency: Autodesk Platform Services (APS), AEC Data Model API, Autodesk Construction Cloud and Common Data Environment (CDE) concepts, Forma, Fusion, Flow, and Autodesk data models. * Enterprise systems of record integration: ERP, PDM, PLM, BIM, project controls, cost, scheduling, and asset management, plus enterprise data platforms (data lakes, warehouses, Lakehouse, analytics platforms). * Constrained enterprise deployment: customer VPCs, private, on-prem, or air-gapped environments, data residency and sovereignty requirements, SSO/SCIM, and formal security review processes. ## Description Autodesk Technical Advisory is seeking a highly technical and customer-focused Embedded Engineer, AI Solutions to join our new Embedded Engineering (forward-deployed engineering) team. You will work directly alongside strategic Autodesk customers to discover high-value workflows, model the data those workflows depend on, and build the AI systems for their Design & Make workflows to enable customer's measurable business outcomes. You will move fluidly from understanding a customer's workflow and architecture to writing code. You will spend your weeks writing production code to build agents and orchestration, retrieval and graph services, connectors and APIs - inside customer environments, alongside customer engineers. What you ship is concrete: AI agents, MCP servers and tools, domain and semantic models, knowledge-graph extensions, retrieval and context-resolution pipelines, integrations into ERP, PLM, BIM, enterprise data platforms, and eval harnesses to move new capabilities into production. Unlike traditional project delivery, Embedded Engineering works deeply inside customer environments, connecting Autodesk technology and data with the rest of the enterprise. Insights developed through customer work will be translated into reusable agents, connectors, reference architectures, engineering patterns, and direct product feedback that shapes Autodesk's platform roadmap., Discover Workflows and Own the Outcome * Partner with customer executives, business leaders, and technical teams to translate transformation objectives, operational KPIs, sustainability commitments, and digital initiatives into measurable, AI-enabled workflow outcomes. * Lead co-innovation with customers to find high-value workflows, decision bottlenecks, data requirements, and integration opportunities, turning ambiguous business challenges into clear technical directions. * Analyze end-to-end workflows across Autodesk and non-Autodesk systems to pinpoint friction, process inefficiency, data gaps, and the places where AI and agents genuinely change the economics. * Own technical execution end-to-end - rapid prototype through architecture, build, integration, testing, governance, production deployment guidance, and adoption - with accountability for customer impact. * Sequence the work and decide what not to build. Make explicit trade-offs between scope, speed, and durability, and defend them to both customer and Autodesk stakeholder Make Data and Context AI-Ready * Assess customer data readiness - fragmented sources, ownership, quality, lineage, metadata, governance, permissioning, and integration constraints - required to support trusted AI and agentic workflows. * Structure and connect design intent, project, product, and asset data, along with requirements, supporting evidence, approvals, decisions, and dependencies, into trusted AI-ready data and context models that enable reasoning, retrieval, and agentic action. * Design ontologies, semantic layers, and knowledge graphs that business users recognize and agents can act on, including entity resolution across source systems, schema mapping, and permission models that survive contact with enterprise reality. Build and Ship Production AI * Design and build production-grade AI-enabled workflows, agents, applications, and integrations that connect Autodesk products and platform capabilities with customers' broader enterprise technology environments. * Build the retrieval and context layer these systems depend on - hybrid search, graph-grounded retrieval, tool interfaces, and MCP servers that expose governed enterprise context to agents safely. * Make quality measurable. Build evaluation harnesses, golden datasets, guardrails, tracing, and observability so that agent behavior is testable before launch and monitored after it, and so that quality regressions and drift are caught in production rather than reported by users. * Apply strong engineering practices across architecture, code quality, security, permissions, reliability, observability, testing, performance, and maintainability, adapting quickly to complex and constrained customer environments. * Define the operating model for what you build - runbooks, ownership, escalation, and cost - and hand it over cleanly to customer teams or Autodesk services so value persists after the engagement ends. Feed the Product and Build Reusable IP * Act as a frontline product discovery mechanism, translating customer learning into prioritized product requirements, platform capability gaps, roadmap recommendations, and opportunities for feature or API acceleration. * Convert customer-specific learning into reusable Autodesk value - reference architectures, agent and MCP patterns, connectors, ontology templates, graph models, reusable code, evals, and technical playbooks - that reduces the cost of every engagement that follows. Build an AI-Native Practice * Design, build, and operationalize the agents, workflows, and automations that make the Embedded Engineering organization itself faster and more effective, measuring success by both customer impact and scalable internal leverage. Continuously look for manual work to replace with agentic systems and reusable internal tooling. * Stay at the frontier. Maintain working knowledge of new model capabilities, agent implementation patterns, and AI product development stacks, and bring what's genuinely useful into practice. * Serve as a trusted technical advisor across the boundary - partnering with customer engineers, architects, and business leaders, and with Autodesk Sales, Customer Success, Technical Advisory, Product, and Platform teams - to unblock delivery and help teams navigate ambiguity, trade-offs, risk, and changing requirements., * AI-native builder practice: routine use of AI coding agents, harnesses, and automation to multiply individual and team throughput, including spec-driven development: writing clear specs and executable acceptance criteria for coding agents, keeping changes small and reviewable, and verifying AI-generated code through automated tests and review. * Production trust and safety: prompt-injection defense, permission-aware retrieval, provenance and audit trails, versioning, staged rollout, monitoring, and incident response. Ideal Candidate * Ships. Has put things into production and can name the metric that moved. * Treats evaluation as part of the product. Assumes agents will fail and builds evaluation, observability, and guardrails before scaling. * Is fluent in both worlds. Understands the customer's business process and the failure modes of non-deterministic systems. * Thinks in reusable patterns. Instinctively asks, "How does this become an asset for the next ten customers?" * Moves fast without leaving a mess . Knows when speed matters, when engineering rigor matters, and how to make that trade-off deliberately.