Applied AI Data Scientist

RELX•New York, NY

1d•Onsite

About The Position

LexisNexis Legal & Professional is hiring an Applied AI Data Scientist to help shape the next generation of AI-powered legal products and experiences. This role involves partnering with internal teams and enterprise stakeholders to design, evaluate, and continuously improve AI capabilities for legal research, drafting, and decision-making. You will work directly with AI engineers, machine learning engineers, and product teams to frame problems, run experiments, design evaluation methodologies, and translate applied research into production-ready AI features. The position is hands-on, focusing on experimentation, evaluation, prototyping, and iterating on AI-powered experiences, requiring a blend of strong applied data science and machine learning fundamentals with practical experience in LLMs, Agentic systems, and AI-native workflows in production settings. Occasional travel to customer sites may be required.

Requirements

6+ years of experience as a Data Scientist, Applied Scientist, Machine Learning Engineer, or related quantitative role, with a track record of shipping data-driven solutions.
Strong foundation in statistics and experimental design (hypothesis testing, A/B testing, causal inference, confidence/uncertainty quantification).
Strong programming skills in Python and its data stack (pandas, NumPy, scikit-learn) plus SQL.
Hands-on experience developing and evaluating LLM-powered or machine learning solutions, ideally in production-oriented settings.
Demonstrated ability to design rigorous evaluation methodologies and metrics for AI/ML systems (offline/online evaluation, error analysis, benchmark construction, quality measurement).
Practical experience with LLM techniques such as prompting, retrieval-augmented generation (RAG), embeddings and semantic search, and structured generation.
Experience with the full modeling lifecycle: data exploration, feature engineering, model training and validation, and monitoring for drift and degradation in production.
Familiarity with modern AI engineering frameworks and tooling such as LangChain, LangGraph, LlamaIndex, OpenAI APIs, Anthropic APIs, or equivalent systems.
Experience working with AI/ML systems and data infrastructure on AWS, Azure, or GCP.
Ability to translate ambiguous business problems into well-scoped, measurable questions and communicate findings clearly to engineering, product, and business stakeholders.
Comfortable working in evolving environments and collaborating across teams to deliver data- and AI-powered features and workflows.

Nice To Haves

Experience in legal technology, enterprise SaaS, compliance, financial services, healthcare, or other regulated industries.
Experience with agentic workflows, multi-step reasoning, or tool-calling systems, including long-running agent design, orchestration of autonomous agents, and the harness and feedback-loop engineering that supports them.
Familiarity with retrieval and ranking optimization, hybrid search, or knowledge graph integration.
Experience with human-in-the-loop evaluation, annotation workflows, or building internal benchmarks.
Experience with AI guardrails, hallucination detection, responsible AI, or grounded/citation-based generation.
Familiarity with fine-tuning or model adaptation techniques.
Experience contributing to AI copilots, AI assistants, or workflow automation systems.
Open-source contributions, technical blogging, conference speaking, or AI/ML community involvement.

Responsibilities

Develop a strong understanding of customer workflows and operational challenges through direct engagement with lawyers, legal operations teams, and subject-matter experts.
Translate ambiguous customer pain points into well-scoped, measurable problem statements.
Collaborate with customers and stakeholders to prototype, validate, and refine AI-powered workflows and user experiences.
Design and run experiments to integrate applied research and emerging techniques (LLMs, RAG, retrieval and ranking, multi-step reasoning, agent patterns, evaluation science) into validated capabilities.
Develop and iterate on LLM-powered approaches such as prompt engineering, retrieval strategies, context management, structured generation, and lightweight agent patterns.
Design and prototype agentic AI systems, including long-running, autonomous agents that plan, call tools, and reason over multiple steps.
Build the necessary infrastructure around agents, including orchestration, state and context management, tool integration, and feedback loops.
Build rapid, runnable prototypes to test ideas, de-risk assumptions, and explore UX and architectural trade-offs.
Analyze model and pipeline behavior, including error analysis, failure modes, and data quality issues, and translate findings into improvements.
Contribute production-oriented code and partner with engineers to harden prototypes for production.
Work with modern AI tooling and frameworks such as LangChain, LangGraph, LlamaIndex, OpenAI SDKs, Google ADK, and/or Anthropic/Claude APIs.
Design rigorous, domain-aware evaluation methodologies for legal AI, covering accuracy, comprehensiveness, citation grounding, hallucination detection, and other quality dimensions.
Define offline and human-in-the-loop evaluation approaches, metrics, and benchmarks.
Build and run evaluation harnesses to compare models, prompts, retrieval strategies, and configurations.
Extend evaluation to long-running, multi-step agents, measuring trajectory quality, tool-use correctness, and agent feedback loop behavior.
Partner with engineers to integrate evaluation, monitoring, and observability into production AI applications.
Balance innovation with practical constraints like latency, cost, reliability, and explainability.
Partner closely with the Applied AI Engineer, machine learning engineers, designers, product managers, legal SMEs, and platform engineering teams.
Communicate clearly with non-data scientists, adapting language to the audience.
Clearly explain how models work, their limitations, and safeguards to build trust with technical and non-technical stakeholders.
Contribute reusable evaluation methods, datasets, and findings back to shared AI platform capabilities and team practices.
Contribute constructively to technical discussions and collaborate effectively across teams.