Database expert Dominik Tomicevic looks at what the space agency’s new people graph pilot reveals about the future of HR technology
Workforce planning is one of HR’s hardest problems to solve. Surfacing skills, finding subject-matter experts, understanding who’s worked on what and building a dynamic picture of the workforce has never been easy, especially when relying on rigid data systems designed for payroll, not intelligence.
Now NASA is trying something different. Its internal People Knowledge Graph pairs graph database technology with large language models (LLMs) to make workforce data more connected, more searchable and far more useful. And while the technology may be complex the lessons for HR leaders are surprisingly simple.
This is a pilot worth watching, not just because it’s AI-powered but because it shows what becomes possible when you rethink how your workforce data is structured, queried and connected.
The challenge
As a chief human resources officer or chief learning officer there are a few things you likely need but still don’t have:
- A fast, scalable way to find subject-matter experts across your organisation
- An easy way to analyse thousands of employee records to identify skill gaps
- Smart tools that match people to the projects where they’ll have the biggest impact
- And one place where all your people data – skills, analytics and L&D – actually connects
HR leaders have long turned to technology for smarter workforce insights, often with underwhelming results. Most people analytics tools still rely on databases that aren’t built to surface complex patterns or navigate networks of expertise.
Graph technology offers a different approach. It structures data not in rows and tables but in relationships – nodes and connections that mirror the way people actually work. Think less spreadsheet, more social graph. Companies like Workday, ADP and Darwinbox are already exploring this model and graph technology is famously behind early versions of Google Search. As global HR analyst Josh Bersin predicted four years ago today’s companies function more like networks, making graph databases a more natural fit than conventional data models.
Like many large organisations, NASA wanted to unlock the expertise hidden across its workforce, making it easier to find the right people with the right skills and connect them to the projects where they could make the biggest impact. But with data scattered across systems and teams, collaboration was harder than it should have been.
What NASA did
NASA’s people analytics team developed a dynamic People Graph, an internal system that maps skills, surfaces subject-matter experts and links people to project needs in real time, supporting smarter collaboration across the agency.
According to its head of analytics for human capital David Meza the system is already being applied across a wide range of HR functions, all the way from guiding upskilling efforts to supporting broader workforce planning. Speaking at a recent webinar Meza described it as an increasingly effective way to connect people, skills and projects across the agency. “Knowledge graphs offer flexibility since you don’t need a full schema upfront,” he notes. “We began with known relationships and expanded as we uncovered more insights in the data.”
Meza runs the graph database on NASA’s private cloud, using AWS (Amazon Web Services) – a secure platform for hosting and managing large-scale data. This forms the backbone of an HR analytics system that maps 18,000 employees as 27,000 ‘nodes’ (people or skills) and 230,000 ‘edges’ (connections between them). The team anticipates this could scale to 500,000 nodes and millions of edges as the NASA People Graph evolves.
NASA has, in fact, been applying graph databases to HR and skills challenges for several years. By pulling data from its personnel data warehouse, employee CVs and machine learning project records, the system can quickly answer previously hard-to-reach questions, like who’s worked on similar projects, who’s skilled in Python, R or COBOL (common programming languages used in data and analytics) or where the agency’s biggest capability gaps lie.
According to NASA’s HR IT team this enables far more sophisticated querying than traditional databases that store data in tables. Graph databases organise information more like a network, which means it can rapidly scan and connect multiple layers of workforce data in seconds, linking people, projects and skills that would otherwise be hard to trace.
In 2024 NASA enhanced the system with large language models, advanced AI tools like ChatGPT that can understand and generate human language, to make it more accessible for non-technical users. The goal was to let people query workforce data as naturally as they would ask ChatGPT only trained and grounded in NASA’s internal knowledge rather than the public internet.
NASA recently added GraphRAG, a tool from Microsoft Research that helps AI systems access more accurate answers by connecting them to structured internal knowledge. Put simply, this lets you access specialised, up-to-date information beyond what a standard ChatGPT can provide, making LLMs smarter, more context-aware and much better at handling complex people analytics and skills forecasting questions. This also dramatically reduces hallucination risk and keeps the system grounded in real, up-to-date workforce data.
The results
The benefits for HR are already visible. The system is helping teams discover relevant experts faster, support organisational intelligence efforts and reduce duplication by linking skills data at the individual level.
At the same time NASA’s leadership is gaining valuable insights into workforce composition and organisational dynamics.
“Where we're seeing it going is to just gain a comprehensive understanding of things like ‘Who is the workforce? Where do we have opportunities to improve efficiency, and better understanding how we maximise our potential?’” says Katharine Knott, people analytics data scientist at NASA.
She adds: “We have a lot of really smart people that work here and we need to work out how to share all of their fantastic knowledge. Having a database we can use to understand who was working on what project, what the important pieces of information that need to be made available for that project is where those are being stored, could therefore be extremely valuable.”
Because skills are now tied to employees at the ID level, with detailed attributes like position title, occupation, pay grade, organisational unit or NASA centre, education and even free-text project descriptions, the organisation has a much fuller view of what its people know and can do.
“It’s still early days at NASA,” says data scientist and data engineer Madison Ostermann, “but the agency is learning, refining and testing across its HR knowledge base.” She says the combination of graph technology, GraphRAG and graph algorithms is proving to be a powerful foundation.
This approach could lead to a general-purpose HR LLM that can help chief people officers quickly access valuable data on employee preferences, learning goals, project interests and more while keeping everything within the safety and specificity of internal data.
Crucially, this approach also tackles the biggest question facing many HR leaders today: can AI meaningfully support goals like skills development, workforce growth and people empowerment?
For HR leaders looking to build skills-based organisations and unlock internal expertise, NASA’s people graph isn’t science fiction but a real-world signal of where the function is heading.
Dominik Tomicevic, pictured below, is the co-founder and CEO of European in-memory database and knowledge graph specialist Memgraph
