Agentic Engineer (all genders)

About Thorit Thorit is one of Europe's leading HubSpot Partners and a leading consulting and technology firm in the DACH region. We combine strategic consulting with hands-on implementation, from CRM architecture to fully automated, AI-powered go-to-market infrastructure. Our clients are enterprise B2B companies who do not just want to adopt a platform, they want a real growth system. Our team is based in Böblingen and Lisbon, and now expanding to Tbilisi, our clients across the DACH region and beyond. Your Role As Agentic Engineer in our Tbilisi hub, you ship at least 2 production-grade multi-agent systems for enterprise clients (delivered via the Lisbon or DACH frontstage) and contribute substantively to Thorit's reusable agent orchestration framework within 12 months. The role is distinct from the AI & Data Engineer: that role builds single agents and the data underneath. You compose agents into systems: orchestration topology, tool ecosystem design, agent memory, human-in-the-loop, long-running loop tracing. You pair daily with the Lisbon Agentic Engineer as your remote tech lead. We are hiring this role because Thorit is a HubSpot Diamond Partner and Anthropic Build Partner, building toward DACH Category Leader for Agentic Revenue Operations by 2028. The AUTOMATE layer of the Agentic Growth Stack is Thorit's central commercial differentiator and is explicitly understaffed. The Tbilisi seat adds capacity behind the Lisbon counterpart at a more flexible cost and legal framework. Note: true multi-agent orchestration experience is thin globally and even thinner in the Tbilisi candidate market, so we welcome strong production AI engineers with 1 to 2 framework-level agent projects, with a 6 to 8 week ramp on Thorit's agent framework during probation. In Phase 1, this role is internal-facing. Your Responsibilities Production Multi-Agent Delivery. Ship 2+ production multi-agent systems used by enterprise clients via the Lisbon or DACH frontstage within 12 months, deployed, signed-off, in active use. Reusable Framework Contribution. Contribute substantively to Thorit's reusable agent orchestration framework: at least 3 framework-level commits or modules merged (orchestrator patterns, tool registry, agent memory, observability hooks). Agent Ownership. Take ownership of at least 1 production agent within 6 months of joining, paired with the Lisbon Agentic Engineer for handover. Quality Bar. Apply Thorit's agent quality bar to every shipped system: documented orchestration-level evals, rollback plan, incident playbook, observability instrumentation. Billable Capacity. Be billable on internal AUTOMATE projects or client engagements via Lisbon/DACH frontstage at 60% of available capacity from month 6. Pre-sales Authority. Contribute to AUTOMATE-layer scoping conversations as the technical authority on the Tbilisi side, via the Lisbon or DACH frontstage. Thought Leadership. Co-author 1+ public-facing technical asset (blog post or webinar) on Thorit's multi-agent architecture approach. What We Offer Competitive compensation with a transparent salary band and a structured bonus model tied to delivery outcomes and team contribution. Remote.com EOR contract with full Georgian compliance: clean payroll in EUR, paid leave, and local social-security alignment. Hardware setup shipped to your home office before day one (laptop, monitor, peripherals). Blinkist Business (4,500+ book summaries) and Babbel (3 months) for continuous learning. Remote Work Program, up to 4 weeks per year from anywhere in the EU, with €500 setup allowance. Cross-site exchange, with a 3 to 5 day on-site visit to Lisbon or Böblingen within your first 90 days, and regular team rituals across all three sites. Tbilisi office access once leased (timeline 2026 to 2027), plus the option to keep working from home or coworking spaces in the meantime. Modern stack: HubSpot, Jira, Confluence, Claude AI, and the freedom to shape how you work. Direct leadership access, flat hierarchies, and visible impact on a hub being built from scratch. How to Apply Send your application through our careers page. No cover letter needed. What we want to see: your CV plus a concrete example of a multi-agent system or a framework-level agentic contribution you shipped end-to-end. Your Profile 3 to 6 years of production engineering with shipped multi-agent systems, OR sophisticated single-agent systems plus 1 to 2 multi-agent framework-level projects. You can articulate when to use one agent vs. many, and the trade-offs. Hands-on experience with at least one major agent framework (LangGraph, LangChain, AutoGen, CrewAI, Semantic Kernel, Pydantic AI, Mastra, or comparable). Deep experience with one framework is acceptable, with a ramp on Thorit's chosen stack. Tool design and composition: you have designed tool schemas that LLMs reliably use, and tuned tool docstrings and parameter naming based on observed model behaviour. Long-running agent loop engineering: you have shipped agents that run more than 10 steps without losing the plot. Retry policies, max-step caps, looping detection, recovery from tool failure are your routine. Agent observability and tracing: you have set up tracing for production agent runs (LangSmith, Helicone, Arize, Langfuse, or rolled-own). You read traces fluently. Production engineering hygiene: maintainable Python or TypeScript with tests, types, CI, deployment. Comfortable in cloud, Docker, basic IaC. Business-fluent English (working language across all sites). Native speaker not required. Tbilisi-based, working under Remote.com EOR contract, with consistent 5-hour CET overlap daily. What Makes You Excel You have implemented agent memory architectures (short-term scratchpad, plus at least one form of long-term memory) in production and know when memory helps and when it adds risk. You have designed human-in-the-loop patterns and you know which decisions should never be fully autonomous. You read token cost and trace duration like an SRE reads CPU. You know when to short-circuit a loop, when to switch models, when to cache. You operate daily with Lisbon and DACH counterparts. Async updates from you remove the need for sync follow-up. You see this role as engineering, not as an entry into AI research. You learn from production deployments, not from notebook prototypes.