As a CHRO or Chief Learning Officer, there are things you wish you had, but likely don’t: a fast way to find internal experts, a simple method to extract skills from thousands of CVs, tools that match people to high-impact projects, and a single place where analytics, skills, and L&D data actually connect.
HR leaders have long turned to technology to solve these challenges, often with mixed results. But a major U.S. agency may be turning the tide. NASA is piloting a “People Knowledge Graph” that’s transforming workforce intelligence.
What makes this use case stand out isn’t just the AI – it’s how AI becomes significantly more effective when paired with tools like search algorithms and knowledge graphs. Together, these technologies are outperforming standalone AI systems, as NASA is now demonstrating.
The problems NASA was trying to solve
NASA’s People Analytics team is leveraging a dedicated People Graph to identify subject matter experts and improve collaboration. It wants to align the right talent with the right initiatives. David Meza, Chief Data Officer for Human Capital, says the system is already supporting a wide range of HR priorities, from upskilling to workforce planning.
In a recent webinar, Meza described the graph as a powerful tool for connecting people, skills, and projects. “Knowledge graphs offer flexibility, since you don’t need a full schema upfront. We began with known relationships and expanded as we uncovered more insights in the data.”
The system runs on NASA’s private AWS network, using Amazon S3 for storage and EC2 for compute. It currently maps 18,000 employees, with a total of 27,000 nodes and 230,000 edges.
GraphRAG provides lift off
NASA has been using graph databases in HR for years, applying techniques like similarity analysis to surface hidden connections across its workforce. This enables powerful queries such as, Who’s worked on similar projects? Who speaks language X or Y? Where are our skill gaps?
According to the team, this richer, more flexible question set marks a significant leap beyond the limited capabilities of traditional relational databases. Graph technology shines in these scenarios, enabling fast, multi-hop traversals across interconnected data points. Rather than relying on rigid tables, it uses nodes and edges, essentially building a network of relationships.
A recent upgrade to this model is GraphRAG, which enhances large language models (LLMs) by adding graph-based retrieval. It gives users access to specialized, current information that standard ChatGPT-style interfaces can’t deliver. It makes LLMs more accurate, context-aware, and capable of handling complex, data-rich queries.
NASA’s team is already seeing results. The finder tool quickly surfaces experts with specific skills, while leadership gains clearer metrics on workforce dynamics. The system is helping reduce duplication, boost efficiency, and unlock deeper collaboration across the agency.
What’s especially promising is how LLMs can now streamline the extraction of insight from unstructured data. NASA’s People Graph is being developed to support natural language queries, like ChatGPT, for workforce planning, data analysis, and agency-wide intelligence.
It uses skills data from NASA’s personnel data warehouse, employee resumes, and internal AI/ML project details. These sources are integrated using GraphQLAlchemy, an open-source Python library and Object Graph Mapper.
A new addition to the people graph: LLMs
Skills are linked to employees as nodes in the graph. Each person is assigned attributes like position title, occupation, pay grade, NASA Centre, education, and free-text project descriptions. LLMs help by recognizing, for example, that “JS” and “JavaScript” refer to the same skill.
The next step, according to NASA Data Scientist Madison Osterman, is using tools such as cosine similarity, vector indexing, and graph-based Retrieval-Augmented Generation (RAG) to make the data easily searchable via a chatbot, not Cypher.
Fellow People Analytics Data Scientist Katherin Knott says the potential is huge. “We haven’t put this into production yet, but where we’re seeing it going is to gain a really comprehensive understanding of things. Who is the workforce? Where do we have opportunities to improve efficiency, understanding how we maximize our workforce?”
It’s still early days, says Osterman, noting that the team is actively learning and testing. Interestingly, the combination of graph technology, GraphRAG, and smart algorithms like pivot search is already proving to be a solid foundation. It is not just about a skills-based People Graph, but potentially for a broader HR-specific LLM. Such a system could incorporate data on employee preferences, learning goals, and ideal project types.
The team believes the graph is highly scalable, capable of supporting hundreds of thousands of employees and billions of nodes.
For any CHRO or CLO still asking, “Do we really need LLMs in HR?” this might be the most compelling answer yet.
Memgraph believes in a future where the problem-solving power of graphs is in the hands of everyone. Frustrated that existing graph database systems were not engineered with performance in mind, the company’s products overcome the limitations of current market solutions with a state-of-the-art, high‑performance, and scalable graph database. Its flagship product, Memgraph, is an in-memory graph database designed for real-time applications such as risk assessment, 360-degree data and network data exploration, Master Data Management (MDM), Bill of Materials (BOM), and supply chain and network logistics. Visit memgraph.com
