Senior Data & Applied Scientist – Ontologies, Knowledge Graphs & AI Research

Description

You will build the semantic layer that lets AI agents act on business data safely, not just read it.

Agents are starting to handle real back-office work such as releasing blocked invoices, re-planning supply and preparing the financial close. Acting safely means knowing what an order, a supplier or a cost center is, how they relate, which rules and permissions apply, and how similar decisions went before. In mission-critical processes, being right 90% of the time falls short.

The company's enterprise platform holds data for many of the world's largest firms. Its knowledge graph covers 452,000 tables, 80,000 CDS views (semantic data models) and 7.3 million fields. Much of the meaning, especially in each customer's extensions, configuration and decision history, is still implicit. A customer in sales, a business partner in finance and an account in CRM can be one company seen three ways, and an agent deciding on a credit limit has to recognize that. This is mainly a knowledge representation problem, not a data cleaning one. Today this layer is mostly modeled by hand for each company. The team believes it can be learned, which underpins the company's autonomous enterprise work and the agents people use in its AI workspace, Joule Work.

The role is a senior research and modeling position. You will own a research thread from an open question, to a method and benchmark, to a prototype a product team adopts. It is not a product analytics role.

The company has operated for over 50 years and serves more than 20 industries, touching 80% of global commerce. The team is a small applied research lab in Palo Alto, minutes from Stanford, that looks a few years ahead, tests emerging technologies and drops ideas quickly when they fail. It has dedicated GPU compute and, in co-innovation projects, access to real data for training and evaluating models.

Responsibilities

• Take an open question from vague framing to a reproducible result, then choose the next step based on the evidence.
• Work closely with leading universities and institutes, such as Stanford HAI.
• Reproduce promising new papers on the company's data within weeks and recommend which ones deserve a larger effort.
• Develop methods to learn and maintain ontologies of business objects, processes, rules and actions, starting from the company's semantic models and extending them with LLMs, schema matching and entity resolution.
• Build graph-augmented retrieval and text-to-query agents (SQL, SPARQL, APIs) for very large schemas, with checks against business rules, access rights and governed actions.
• Apply graph neural networks and relational foundation models, with knowledge graph context, to prediction, anomaly detection and what-if analysis.
• Create datasets, benchmarks and evaluation harnesses that show whether an approach works on realistic enterprise data and explain why, using synthetic or privacy-preserving data when real data cannot be shared.
• Write clean Python that others can build on, co-build prototypes with product and engineering teams, publish papers or open-source code, and mentor interns.

Requirements

• Required: 5+ years of hands-on data science, machine learning or applied research, taking projects from an open question to a measurable result. Or a PhD plus 2+ years in the same work.
• Required: an MS or PhD in computer science, machine learning, statistics, mathematics or a related field, or equivalent practical experience.
• Required: deep hands-on skill in at least one of these areas, with a willingness to learn the others: knowledge graphs, ontologies or semantic layers (RDF, OWL, SHACL, SPARQL, or property graphs with Cypher or similar); entity resolution, schema matching or data integration; machine learning on relational, graph or tabular data, including GNNs; LLM systems over structured data (text-to-SQL, schema linking, API discovery, GraphRAG, agents, evaluation).
• Required: solid grounding in machine learning and statistics, with disciplined experiments that include strong baselines, ablations, leakage-safe splits, error analysis and honest reporting of failed approaches.
• Required: strong Python and SQL, and PyTorch or a similar framework, comfortable with large, messy, multi-table data.
• Required: concrete examples of turning a vague question into an experiment that gave a clear answer within weeks.
• Required: concrete examples of writing up a negative or mixed result so clearly that others acted on it, and of explaining your work concisely in conversation and in a one-page summary to a scientist or executive.
• Preferred: publications at venues such as NeurIPS, ICML, ICLR, KDD, ACL/EMNLP, SIGMOD/VLDB, ISWC or WWW, or contributions to open-source projects.
• Preferred: graph databases, vector search, or large-scale data processing such as Spark.
• Preferred: LLM-assisted ontology learning or knowledge graph construction.
• Preferred: LLM fine-tuning or post-training, or MCP.
• Preferred: process mining, causal inference or forecasting, and familiarity with enterprise processes in finance, procurement or supply chain.
• Preferred: experience with the company's software is a plus, but not required.
• Candidates who match only some of these points are still encouraged to apply.

Conditions

The role is hybrid, with three days a week in the Palo Alto office and travel of up to 10%.

This is a senior individual contributor role, with peers in Senior Research or Data Scientist positions at other Bay Area tech companies.

The salary range is listed in the posting. Final pay falls within that range and depends on education, skills, experience, location and scope of the role, as decided through selection.

Variable incentives may be offered under the relevant plans. Any incentive is target-based, and actual payouts depend on company and individual performance.

Use of AI in recruiting is governed by the company's ethical guidelines, and violating them may lead to disqualification.

Applicant information is kept confidential under equal employment opportunity guidelines.

Benefits

• Ongoing learning and skill growth, with a team that supports your development.
• A North America benefits summary is published by the company, covering eligibility.
• Permanent employees may take part in the employee referral program.