Senior AI Engineer

Senior AI Engineer — NetSpeek

NetSpeek is the agentic control plane for enterprise physical infrastructure . We govern how AI agents reason about, decide on, and execute actions across enterprise endpoints. Our reasoning and execution layer — Lena — sits in customer production environments, where reliability, observability, and auditability are non-negotiable.

The Senior AI Engineer owns Lena's reasoning layer end-to-end: retrieval, grounding, evaluation, and the boundary between AI suggestions and governed actions.

What you'll work on
- Designing and improving RAG pipelines for grounding Lena's diagnostic reasoning in structured operational telemetry, device state, and product documentation.
- Building evaluation harnesses that measure groundedness, hallucination, refusal calibration, and action accuracy on every release.
- Setting the boundary between Lena's probabilistic reasoning and the platform's deterministic action layer — what she's allowed to do, when, and under what audit.
- Owning AI cost and latency budgets per workflow.
- Partnering with backend (.NET) and platform engineers to land changes safely.

You're a fit if
- You have 5+ years of ML / applied AI engineering experience.
- You've built and shipped production LLM systems (RAG, agents, structured outputs, evaluations) at a B2B SaaS company.
- You've owned a production RAG system end-to-end.
- You've built evaluation pipelines that ran on every release and caught real regressions.
- You've worked at a growth-stage AI-native SaaS company where AI was the primary product.

You probably aren't a fit if
- Your AI exposure stops at experimentation or coursework.
- You haven't deployed AI systems to customer production environments.
- You want a process-heavy environment where decisions go through committees.

How to apply

We prefer applications through our hiring site: https://build.netspeek.ai/roles/senior-ai-engineer

There are three paths there, and we prioritize them in this order:

1. Incident Lab / Case Study —work through a real scenario for this role and send your reasoning with the application. Read first. The scenario gives us the most signal on how you actually think.
2. Field Note —a single written take instead of the scenario. Read second. Faster than the lab, still a strong signal.
3. Standard application —resume and the basics through the site's apply form. Read third. Applying directly through Workable below also works, and we'll still read it. The site's three paths exist because they give us better context for the read.

Either path is read by a human on our hiring team. No AI scoring, no auto-rejection.
Read the Engineering Handbook and How We Evaluate before applying.

After an offer

We run standard pre-employment checks before your start date: identity verification, right-to-work confirmation, employment verification, and (where lawful and role-relevant) a criminal record check. We don't run credit checks or online reputation scoring.