Employment Type: Full-time
Hybrid is optional.
About Us
We’re building an advanced “attorney assistant” system that leverages cutting-edge Large Language Models (LLMs) for EU law. Our goal is to create an AI-driven platform capable of understanding and referencing legal statutes, court decisions, and other documentation.
We’re looking for a Senior NLP/LLM Engineer who will spearhead the entire process—from ingesting and chunking legal PDFs to fine-tuning large models and deploying retrieval-augmented pipelines. You’ll have direct access to powerful GPU infrastructure, including NVIDIA Spark technology.
Key Responsibilities
- Data Pipeline & Preparation:
- Extract and clean text from PDFs (legal statutes, case law, etc.)
- Chunk documents and manage large-scale text corpora
- Embeddings & Vector Databases:
- Select and implement multilingual embedding models
- Set up and optimize vector databases (FAISS, Chroma, Weaviate, etc.)
- Fine-tune retrieval parameters for best-in-class results
- LLM Fine-Tuning & Inference:
- Perform parameter-efficient fine-tuning (LoRA / PEFT) on specialized legal datasets
- Handle large checkpoints on GPU systems (memory constraints, distributed training as needed)
- RAG Pipeline Development:
- Integrate fine-tuned LLMs with retrieval-augmented generation (RAG) workflows
- Architect end-to-end solutions for legal Q&A and summarization in
- Deployment & MLOps:
- Containerize and deploy the final pipeline (e.g., Docker, FastAPI/Flask)
- Implement continuous improvement, versioning, and updates to legal data
- Collaboration & Knowledge Sharing:
- Work closely with cross-functional teams (developers, product, legal advisors)
- Document progress and mentor junior engineers (if applicable)
Requirements
- 3+ years of hands-on NLP experience (transformer-based models, text processing, etc.)
- Proven track record using Hugging Face Transformers, PyTorch, TensorFlow (or similar)
- Experience with vector embeddings and RAG (retrieval-augmented generation)
- Understanding of LoRA/PEFT or other parameter-efficient fine-tuning methods
- Familiarity with GPU environments (CUDA, performance tuning, multi-GPU usage)
- Strong Python skills (data wrangling, automation scripts, etc.)
- Bonus: Knowledge of Croatian or other Slavic languages; previous work with legal or domain-specific text corpora
What We Offer
- Competitive Compensation: See salary range below
- Challenging Projects: Cutting-edge legal AI in a specialized domain
- Modern Hardware: High-end GPU servers (NVIDIA Spark) for large-scale experimentation
- Flexible Work Arrangements: Remote, hybrid, or in-office
- Growth Opportunities: Chance to lead technical initiatives and define new AI-driven products
- Supportive Environment: A team that values innovation, learning, and collaboration
Salary Range
- 3,000 – 4,500 EUR (net) per month
Depending on experience, scope of responsibilities, and exact skillset. Please apply if you posess the needed skillset.
