Lead Specialist, AI Scientist

Role Overview 

We are seeking a Lead Specialist, Applied AI Systems Engineer, to design, build, and continuously improve the intelligent systems that support Pearson’s work developing the next generation of K–12 assessment and learning. 

This role sits within the Measurement & Learning team and operates at the intersection of Content Development, Learning Science, Technology, AI Science, and Psychometrics, partnering closely with software engineers and the AI Data & Knowledge Architect to turn assessment and measurement expertise into testable AI capabilities within our assessment authoring system, which supports formative and summative assessment for Pearson and its partners. 

This hands-on role joins in the early phases of this work, improving and extending complete AI systems built on a shared knowledge foundation, rather than focusing on prompts or chatbots. The work spans predictive machine learning, statistical methods, retrieval, graph-based approaches, generative and agentic AI, and deterministic software, chosen for reliability rather than novelty. 

We are particularly interested in innovators who have built AI or machine-learning systems whose outputs must be rigorously evaluated against ground truth, expert judgment, or reliable evidence, such as scientific, healthcare, financial, cybersecurity, legal, or assessment applications. Candidates whose experience is mainly in conversational or customer-service AI should also show substantial experience evaluating systems against defined performance standards. Experience in education, learning, or assessment is strongly preferred when it involves the systems beneath products, such as domain and skill models, learner models, standards and competency frameworks as structured data, or adaptive and recommendation engines, rather than standalone tutoring, chatbot, or content-generation applications. 

End-to-End AI Systems Design 

• Design complete, modular AI systems spanning knowledge and data inputs, retrieval, models, agents, tools, deterministic services, human review, evaluation, and integration 
• Break complex problems into the right combination of predictive ML, statistical methods, generative models, retrieval, graph reasoning, deterministic software, and human decision points 
• Build working systems and reference implementations, and partner with software engineers to bring them into the assessment authoring system 
• Design architectures that can evolve as models, data, domain knowledge, and evaluation evidence change 

Agentic & Knowledge-Grounded AI 

• Design and implement agentic and multi-step workflows using models, tools, retrieval, state, structured outputs, and human decision points 
• Build knowledge-grounded systems that use graph, semantic, vector, and hybrid retrieval as appropriate 
• Co-design knowledge and graph architecture with the AI Data & Knowledge Architect, contributing substantively rather than treating the knowledge layer as a black box 
• Develop AI capabilities for content creation and revision, verification, alignment across frameworks, guidance, and analytic insights 

Evaluation-Driven Development 

• Make evaluation part of how systems are built, not a final step 
• Build and maintain evaluation benches, including expert-adjudicated datasets, challenge sets, regression suites, and protected holdout data 
• Partner with psychometricians and content experts to turn domain expectations into measurable criteria 
• Analyze failure modes, critical errors, and performance across meaningful data and task slices, not only aggregate metrics 
• Monitor deployed systems for drift and emerging failures, and use findings to drive improvement 

Machine Learning, Model Adaptation & Data 

• Develop predictive models, classifiers, and ranking methods where they outperform generative AI 
• Evaluate tradeoffs among prompting, retrieval, structured knowledge, model selection, and model adaptation, without locking the architecture to a single provider 
• Apply parameter-efficient fine-tuning and other adaptation methods when evaluation evidence shows they add value 
• Turn expert judgments and adjudicated examples into versioned evaluation and training assets, keeping training data strictly separate from protected evaluation data 
• Design synthetic-data approaches with rigorous quality controls, never treating generated data as ground truth by default 

Innovation, Judgment & Responsible AI 

• Work with a high degree of autonomy on problems that have no established solutions 
• Prototype quickly, test assumptions empirically, and pivot when evidence warrants 
• Build AI that is accurate, transparent in its reasoning, auditable, and designed to support rather than supplant human expertise 
• Uphold responsible AI and data practices, including privacy, security, fairness, and appropriate use of student and content data 

Cross-Functional Collaboration 

• Partner with Technology teams and software engineers to integrate AI capabilities into the assessment authoring system 
• Collaborate with AI Science colleagues and the AI Data & Knowledge Architect on shared architecture and methods 
• Work with psychometricians and learning scientists to translate measurement and learning requirements into evaluation criteria 
• Partner with content development teams to understand workflows, quality requirements, edge cases, and human-review needs 
• Communicate findings, tradeoffs, and limitations clearly to technical and non-technical audiences 

Required Qualifications 

• Advanced degree in Computer Science, Artificial Intelligence, Machine Learning, Statistics, or a related field, or equivalent experience 
• Experience designing and building complete AI or machine-learning systems end to end, not only isolated models or prompts 
• Strong breadth across predictive ML, statistical learning, and LLM-based, agentic, and retrieval-grounded systems, with judgment about when each fits 
• Demonstrated experience evaluating probabilistic systems against ground truth or expert judgment, including evaluation datasets, error analysis, and regression testing 
• Experience with knowledge graphs and knowledge-grounded AI sufficient to contribute to architecture decisions 
• Strong Python proficiency, and experience with TypeScript or the ability to become productive quickly in a TypeScript codebase 
• Track record of innovating beyond established practice, with sound technical judgment and high autonomy 
• Ability to collaborate closely with engineers and domain experts, translating expertise into requirements and evaluation criteria 

Preferred Qualifications 

• Experience building the knowledge and inference layers of education, learning, or assessment systems, such as skill or competency graphs, learner or domain models, standards alignment engines, adaptive or recommendation logic, or the inference behind learner insight reporting 
• Experience building AI or ML systems in evidence-sensitive domains such as scientific, healthcare, financial, legal, or assessment applications 
• Hands-on experience with parameter-efficient fine-tuning or other model-adaptation methods 
• Experience designing, filtering, and validating synthetic data for evaluation or model improvement 
• Experience building expert-annotated datasets, adjudication protocols, or evaluation platforms 
• Experience with graph-grounded agents, hybrid retrieval, or other knowledge-intensive AI applications 
• Familiarity with ontologies, semantic models, or hybrid neural-symbolic architectures 
• Experience with calibration, uncertainty estimation, drift detection, or model monitoring 
• Experience with TypeScript, AWS, PostgreSQL, and graph data stores; our environment also uses Go, Python, and ML lifecycle tools such as MLflow 
• Familiarity with educational measurement, psychometrics, or learning science 

Design Philosophy 

• Complete systems over isolated models 
• Evaluation as part of engineering, not an afterthought 
• Innovation at the edge of current practice, grounded in evidence 
• Transparent, auditable, and accurate AI 
• AI that supports rather than supplants human expertise 
• The right method for the problem, not the newest one 

Why This Role Matters 

This role will build the intelligent core of an industry-defining assessment system. The goal is not convincing AI output but systems whose quality can be measured, explained, and continuously improved. By inventing new approaches and proving they work, this engineer will help define what intelligent formative and summative assessment can become. 

Compensation at Pearson is influenced by factors including skill set, experience, and location. 

The full-time salary range for this role is $140,000 – $160,000. This position is eligible to participate in an annual incentive program. Information on benefits can be found here.

Applications will be accepted through 20th November 2026. This window may be extended depending on business needs.

Apply today and imagine the impact you can make. 
#LI-P1

#LI-Remote