Note: The job is a remote job and is open to candidates in USA. The company supports legal, tax, and accounting professionals with trusted information, software, and AI-powered products. The Lead Applied Scientist will build and deploy document understanding systems, including semantic chunking, information extraction, classification, and knowledge graph pipelines, while partnering with engineering and product teams to move models into reliable production systems.
Responsibilities
- Design and deploy semantic chunking models for lengthy, complex documents, going beyond fixed-size or paragraph-based approaches
- Build document enrichment systems that classify content against legal, tax, accounting and customer-defined taxonomies
- Develop LLM-based information extraction and knowledge graph pipelines, identifying and linking entities, citations, concepts and relationships across documents
- Work on hierarchical and multi-label classification for complex document structures
- Develop systems for entity recognition, entity linking, relation extraction and citation parsing
- Apply knowledge distillation and model compression to develop smaller, production-ready models where latency and cost matter
- Design synthetic data generation and annotation workflows to support model development and evaluation
- Build component-level and end-to-end evaluation frameworks, combining expert annotation, synthetic data and automated evaluation
- Make technical decisions around modelling, architecture, chunking, retrieval and knowledge extraction approaches
- Partner with engineering to take models into production and ensure they perform reliably at scale
- Mentor other applied scientists and ML practitioners and help raise the technical bar across the team
- Provide technical input into the wider AI strategy and roadmap
Skills
- You'll have a strong background in applied NLP, document understanding or information extraction , with experience taking machine learning systems from research and experimentation into production
- • Document intelligence and complex document structures
- • Semantic or structure-aware chunking
- • Hierarchical and multi-label classification
- • Information extraction from unstructured content
- • NER, entity linking and relation extraction
- • Citation parsing and knowledge graph construction
- • LLM-based extraction and post-training
- • RAG, retrieval or question-answering over large document collections
- • Knowledge distillation, model compression or SLM deployment
- • Synthetic data generation and annotation workflows
- • Designing evaluation frameworks for NLP or document understanding systems
- • You'll also be comfortable working in Python , with production experience using technologies such as PyTorch and Hugging Face Transformers
- A PhD or equivalent research background in Computer Science, AI, NLP or a related field is preferred, alongside a strong track record of applied research and production delivery
Company Overview