AI Engineer

Full-timeOslo, Norway · HybridCompetitive salary

You will work on the core of NeuroStrike: multi-agent orchestration, tool use, model evaluation and fine-tuning of open-weight models that run on European infrastructure. Your work directly decides how many real vulnerabilities we find and how reliably we prove them.

What you'll do

Design and improve agentic workflows: planning, tool calling, memory and context management
Build evaluation harnesses and benchmarks that measure real exploitation success
Fine-tune and serve open-weight models (SFT, preference and RL-based methods)
Optimise cost, latency and reliability across multiple model providers
Turn engine traces into high-quality training data

What we're looking for

You must be willing and eligible to be authorised for classified information at BEGRENSET / NATO RESTRICTED level, and to undergo security clearance (sikkerhetsklarering) under the Norwegian Security Act if a customer assignment requires it. This is required due to the nature of our customers.
Professional experience building production systems on top of LLMs
Strong Python and/or TypeScript skills
Hands-on experience with evaluation — you measure before you claim an improvement
Understanding of model limitations: hallucination, prompt injection and tool-use failure modes
Professional working proficiency in English

Nice to have

Experience fine-tuning models (LoRA/QLoRA, Axolotl, TRL or similar)
Experience self-hosting models (vLLM, SGLang, Ollama)
Background in security or interest in offensive security
Publications or open-source contributions in ML

What we offer

Competitive salary and terms
A central office at Aker Brygge, Oslo, with hybrid working