Data Scientist

1. Applied Machine Learning

• •Problem Framing: Translate ambiguous client requirements into well-posed machine learning problems with defined success criteria and realistic scope.
• •Modelling: Select, implement, and tune models appropriate to the problem and the data available, resisting unnecessary complexity where a simpler approach performs.
• •Statistical Rigour: Apply sound experimental design, validation strategy, and error analysis, including investigation of bias, data leakage, and distribution shift.
• •Data Exploration and Feature Engineering: Conduct thorough exploration of structured and unstructured data, and perform the cleaning, transformation, and feature construction needed to make it modellable.
• •2. NLP and Language Technologies
• •Model Development: Design, fine-tune, and evaluate NLP models for tasks including classification, information extraction, entity recognition, summarisation, and semantic similarity.
• •Information Retrieval: Build and evaluate retrieval systems spanning sparse, dense, and hybrid approaches, including embedding model selection, chunking strategy, reranking, and query understanding.
• •Generative AI: Apply large language models to client problems through prompting, structured output, RAG architectures, and fine-tuning where warranted, with a clear understanding of each failure mode.
• •Evaluation Design: Define task-appropriate metrics and build evaluation harnesses for systems where ground truth is expensive, ambiguous, or subjective.
• •3. Knowledge Representation and Graph-Based Approaches (desirable)
• •Ontology and Schema Design: Contribute to the taxonomies, ontologies, and graph schemas used to structure client domain knowledge.
• •Graph Construction from Text: Apply entity recognition, entity linking, and relation extraction to build knowledge graphs from unstructured sources.
• •Graph-Augmented Retrieval and Reasoning: Explore graph-based approaches to retrieval and reasoning where relational structure adds value beyond vector search alone.
• •4. Delivery and Engineering Quality
• •Production-Ready Code: Write clean, well-tested, maintainable Python in line with Attercop's coding standards, contributing to shared codebases and reusable components.
• •Hand-Off and Integration: Work closely with AI Engineers to move models and pipelines into client systems, providing the documentation, tests, and interfaces required for a clean handover.
• •Tooling: Assist in evaluating and adopting new tools, frameworks, and techniques that enhance the team's capabilities and the quality of client deliverables.
• •Ownership: Take responsibility for the quality of your own output, seek peer review proactively, and escalate technical risks or blockers promptly and constructively.
• •5. Research and Innovation
• •Horizon Scanning: Stay current with developments in NLP, information retrieval, and generative AI, bringing relevant ideas and techniques to the attention of the wider team.
• •Prototyping: Contribute to internal R&D efforts, prototyping new approaches and helping to assess their viability for client applications.
• •Documentation: Record research findings, methodologies, and implementation details clearly, sharing knowledge with colleagues and, where appropriate, external audiences.