about valkyria
Valkyria is a defense-focused software engineering company, built in Ukraine, delivering AI solutions for DefTech companies worldwide. We combine 10+ years of AI engineering expertise with 4+ years of real-world combat testing in Ukraine – covering the full cycle from data collection and annotation to algorithm development, sensor integration and deployment on the edge.
RESPONSIBILITIES
Design and build LLM-based applications: RAG pipelines, agents, structured information extraction.
Fine-tune and adapt open-source LLMs for domain-specific defense tasks, including Ukrainian-language data.
Deploy LLMs in secure, on-premise and air-gapped environments.
Optimize inference (quantization, batching, serving) for constrained hardware.
Build evaluation frameworks to measure accuracy, hallucinations and robustness.
Collaborate with domain experts to translate operational needs into technical solutions.
rEQUIREMENTS
Hands-on experience in MilTech / Defense projects (e.g., UAV/UGV, ISR, EW, C2, targeting, sensor systems) – required.
3+ years in ML/NLP engineering, including 1+ year of hands-on work with LLMs.
Strong Python; experience with PyTorch and the Hugging Face ecosystem.
Experience building RAG systems (vector databases, embeddings, retrieval optimization).
Experience with fine-tuning (LoRA/QLoRA, SFT) and self-hosted serving (vLLM, TGI, Ollama or similar).
English – Upper-Intermediate or higher.
NICE TO HAVE
Experience with multimodal (vision-language) models.
Experience with OSINT / intelligence data processing.
Knowledge of LLM security (prompt injection, data leakage).
Military service background.
WHAT WE OFFER
Work on mission-critical defense technology with direct impact on Ukraine’s defense capabilities.
Fast feedback loop from real operational use — your work is tested in the field, not just in the lab.
A team that combines strong engineering expertise with real military experience.
Ownership of meaningful technical decisions and room for professional growth.
Competitive compensation, flexible work format.
.png)