You will be part of the startup's leadership team, reporting to the CEO and will work closely with our COO/Product Leader. In building the technical team, you will hire deliberately; each addition should answer a clear capacity or expertise requirement and enhance our AI-first culture.

What You'll Do

• Own the control plane. Take ownership of the shared-services architecture the platform runs on. Complete the model-abstraction layer so nothing is wired to a single model or cloud. Complete the telemetry so usage and cost are measured everywhere. Structure services so parts can be rewritten quickly.
• Own the agent architecture. Design and build the orchestration model behind the platform, and the Agentic Studio where new agents are defined and tested: specialized sub-agents with narrow scope, human review gates, and an audit path behind every output.
• Make evaluation the operating discipline. Build and run the harnesses that measure faithfulness, alignment with expert content, instruction-following, latency, and cost, and set the standard for what ships. In our domain recall failures cost more than precision failures; the system must reflect that.
• Build the expert knowledge system. Turn the judgment of our expert panel into agent behavior, with a defined order of precedence among customer policy, regulatory law, and commentary. Build the feedback loop that improves the agents with every engagement.
• Engineer for data sovereignty. Own the architecture that lets one platform run in multiple sovereign environments: in-country cloud regions, tenant-isolated data, and the replacement or self-hosting of third-party dependencies where residency requires it — a decisive purchasing criterion in our first market.
• Control cost and reliability. Route work to the right model for the task rather than defaulting to frontier models, including fine-tuned open-weight models as appropriate. Own latency, cost attribution, and the guardrails that keep probabilistic systems predictable.
• Automate the engineering itself. Extend our AI-first engineering practice: coding agents in the development loop, QA harnesses that exercise agents with simulated users, and instrumentation in place of manual review.
• Certifications. Establish and implement architecture refinements to achieve SOC 2 and related certifications critical to enterprise sales.
• Lead a deliberately small team. Hire to a high standard, building flexibility and nimbleness into the organization from day one. Grow the team's output faster than its headcount.

What You Have

• Agentic systems in production. You have shipped multi-step agentic systems with tool use, sub-agent orchestration, persistent memory, and human-in-the-loop review. You can describe specific failure modes and the instrumentation you used to resolve them.
• Evaluation as a discipline. You have designed and operated evaluation harnesses across model versions and prompt changes. We treat this as the clearest indicator of production experience.
• Current LLM engineering practice. You have spent the last two years building with generative AI in commercial products, not research settings. System-level prompt design, retrieval-augmented generation, structured generation, fine-tuning trade-offs, and have a working view of the current model landscape and its costs.
• You support what you ship. You have supported the systems you built through customer feedback, changing requirements, and production incident management. Your record demonstrates sustained ownership rather than a sequence of short engagements.
• Safety and security engineering. Prompt injection mitigation, plan-then-execute patterns, output validation, tool-use restrictions, and policy enforcement.
• Neutral technologist. Multi-cloud and multi-region by experience. You hold no fixed preference for any model or cloud provider. You select what the problem and the region require, and your architectures keep those selections flexible.
• Enterprise production rigor. Python. Multi-tenant architecture, CI/CD, observability, and cost attribution.
• Leadership at a small-team scale. You have led, or been the senior engineer on, teams of five to fifteen. Prior leadership experience qualifies; so does a senior engineer prepared for a first leadership position. Recent experience managing a large organization, rather than building, is not ideal.

Nice to Have

• Experience in a regulated industry: privacy, financial services, healthcare, or legal technology.
• Experience deploying and fine-tuning open-weight models end to end.
• Familiarity with data privacy frameworks such as the GDPR.
• Arabic language capability or working experience in the Gulf region.

Logistics

Remote, UK or EU preferred, but open to other home locations, with working hours overlapping European and Gulf time zones. Occasional travel to customer sites and team gatherings. Competitive base salary for an early-stage seed series startup and founding-team equity.

The Process

Initial conversation with the CEO, followed by meetings with the COO and members of the technical team. References. Offer. Targeting three weeks from first conversation to decision.

Pyxos is an equal opportunity employer. We hire on the basis of capability, judgment, and fit with what we are building. We welcome applications from candidates of all backgrounds and will make reasonable adjustments through the interview process on request.