Agents, assistants, MCP integrations, and automation for jobs people already do. No chatbot looking for a reason to exist. The mandate.
Own the outcome, stay close to the implementation, and improve the system around the work. Build agentic systems around real workflows, tools, and data, with measurable value, explicit human ownership, and enough evaluation to know when the system is useful or wrong. This is software engineering with probabilistic components, not a prompt‑writing role. Security, observability, failure handling, and product adoption remain first‑class concerns. Impact Material Agency High Management Context, then space Bureaucracy Low What this role owns.
OWN-01 Map repeated work before automating it, then choose where deterministic software, LLMs, or human judgement should own each step. OWN-02 Build agents and assistants that use APIs, MCP servers, internal tools, and governed data sources safely. OWN-03 Create evaluation datasets, quality gates, tracing, feedback loops, and production observability for model behaviour. OWN-04 Design permission boundaries, auditability, data protection, and human approval for consequential actions. OWN-05 Work with users and domain experts to ship, measure, and continuously improve the complete workflow. What this needs.
Strong Python or TypeScript software engineering and experience integrating production APIs and data systems. Hands-on work with LLM APIs, agent SDKs, tool calling, retrieval, MCP, and structured outputs. Experience evaluating model quality, latency, reliability, cost, and failure modes beyond demo prompts. A practical understanding of identity, permissions, sensitive data, and human-in-the-loop control. Product judgement: the ability to reject AI when a script, form, search index, or workflow change is better. What success looks like.
The system removes measurable work without hiding ownership or creating a new review bottleneck. Tool use is scoped, auditable, and recoverable when models or dependencies fail. Evaluation and telemetry catch regressions before users have to explain that the assistant became worse. Users adopt the workflow because it fits their job, not because the interface says AI. Top-quartile ranges for this scope. Ranges are base salary; equity and bonus may be additional. The cash/stock split is negotiable. EMEA
€175k–€245k
Base salary Top-quartile EMEA target US remote
$235k–$300k
Base salary Remote US target US major hubs
$290k–$350k
Base salary San Francisco, Seattle, New York, and comparable markets Who you need to be
At least 7+ years of software engineering experience, including hands-on work with production AI or agentic systems. Built agents and assistants connected to real tools, APIs, workflows, and governed data. Strong Python or TypeScript skills with experience in LLM APIs, MCP, agent SDKs, and structured tool calling. Implemented evaluations, tracing, observability, security boundaries, and human ownership for AI workflows. Created automation that removed measurable developer, field, or operational toil. Looks compatible.
#J-18808-Ljbffr
AI Engineer, Developer Automation in Amsterdam werkgever: Mbianchi
Mbianchi biedt een stimulerende werkomgeving voor Senior Staff Site Reliability Engineers, waar je de kans krijgt om impact te maken op de servicegezondheid en operationele processen in cloud-native omgevingen. Met een focus op samenwerking en innovatie, moedigen we een cultuur aan waarin blameless incident reviews en continue verbetering centraal staan, wat bijdraagt aan zowel persoonlijke als professionele groei. Bovendien profiteer je van flexibele werkuren en de mogelijkheid om volledig remote te werken, waardoor je werk-privébalans optimaal blijft.
