AWS Public Sector Blog
Amplifying builder impact: How Kiro transformed cloud operations and security at Loyola Marymount University
How long would it take to update the Python runtime versions across 500 AWS Lambda functions running across an organization’s cloud environment? For the two-person cloud team at Loyola Marymount University (LMU), it used to take two months to inventory, scan, remediate, test, validate, and push to production. Today, it takes just half a day thanks to Kiro—a spec-driven artificial intelligence (AI) development tool from Amazon Web Services (AWS).
Unlike traditional AI coding assistants that require repeated explanations, Kiro’s spec-driven approach allows teams to encode their standards, requirements, and institutional knowledge into reusable specification files. This means Kiro consistently follows LMU’s established patterns across every project, eliminating the need to re-explain requirements with each new task.
Building a cloud footprint from the ground up
Founded in 1865, LMU is a private Jesuit and Marymount research university in Los Angeles, California, serving more than 10,000 undergraduate and graduate students. Over the past decade, Rob Larmon, LMU’s manager of cloud and automation, has worked with a single contractor to evolve the institution’s cloud footprint from a first workload managing disaster recovery backup in AWS, to a full data center migration. But as their AWS environment grew to dozens of AWS accounts, they increasingly relied on strategic consultants and professional services for the most complicated projects.
When Larmon saw the announcement for Kiro’s pilot program in July 2025, he recognized the opportunity. “I could take all our standards, requirements, and the way we do things at LMU and put them into spec files,” he said. Unlike generic AI tools that would forget LMU’s requirements with every conversation, Kiro’s project spaces could encode institutional knowledge once and reuse it across every project. Larmon signed up for the waitlist on the spot.
The challenge: Scaling cloud operations with limited resources
Larmon spent the first few months getting acquainted with Kiro’s capabilities. He started with a one-off project: updating a set of generic Shibboleth cloud formation templates that did not work out-of-the-box in the LMU environment. Within an hour, Kiro delivered templates that met LMU’s tagging, security best practices, and coding standards. “These were templates I would have been proud to build myself,” said Larmon. The success of this project spurred him to try something more ambitious.
Larmon thought about all the Lambda functions running at LMU with out-of-date Python code, some developed in-house and others delivered by vendors. There were more than 500 to update. Larmon asked Kiro to “build a Lambda scanner, tester, and remediator tool tool that goes across all AWS accounts, and compiles an executive summary, including whether the Lambda functions were built from an AWS CloudFormation Template, manually, or from another source.” Kiro helped him get an accurate inventory, develop a plan, and take action.
In four to six hours, Kiro built the scanner-remediator-tester and baked in LMU’s governance structure for repeated use. The Kiro-built tool ran for half a day, inventoried the estate, built the necessary updates, ran checks, created an executive report, and tested every function it upgraded. Once the tests passed, it flagged which updates were ready for Larmon to push to production. “The drudgery Kiro saved,” Larmon remarked. “I would not have attempted to do this at all before.” Without Kiro, recurring runtime updates would have required sustained manual coordination and testing that his small team simply didn’t have bandwidth for.
Kiro delivers consistency, compliance, and accuracy
As Larmon’s experience with Kiro deepened, he began using it to enforce governance and standards while enabling repeatability and scalability. On top of AWS Control Tower, Larmon built a superset of information for Kiro to reference. Now, before Kiro starts any project, it already references LMU’s standards for development, documentation, tagging, accounts, business rules, and more.
The results have been significant. “LMU has consistent, enterprise-grade deployments across all AWS accounts through Kiro’s steering and standards model. This eliminated much of the drift, inconsistency, and re-work LMU used to fight,” said Larmon.
Kiro delivers automatic generation of secure, compliant infrastructure templates (AWS CloudFormation, Terraform/OpenTofu, and AWS Cloud Development Kit) that reflect LMU’s governance rules without needing deep specialist intervention every time. Kiro also provides cross-system alignment with Confluence and Jira, “giving us documentation, diagrams, and work tracking that reflect the actual state of our cloud environment, not whatever was last manually updated,” said Larmon.
Expanding Kiro across LMU’s technical teams
Larmon has seen Kiro deliver cloud solutions for his projects four to 10 times faster than previously, and he wants other teams to experience similar productivity gains. After proving its impact in cloud operations, Larmon has begun selectively applying Kiro to other ITS workflows, exploring integrations with ServiceNow and other enterprise tools to reduce manual effort and improve consistency. With continued validation within ITS, he sees potential to extend these capabilities to support broader campus research and innovation initiatives.
Larmon anticipates broader gains in problem-solving, innovation, and creativity as his colleagues start to explore the tool’s possibilities. On a larger scale, he expects more small teams at LMU to operate with enterprise-level maturity. Thanks to automated documentation and Kiro’s enforcement of governance and other standards, he expects a reduction in dependence on “tribal knowledge.” Training student workers to use Kiro can help LMU researchers accelerate their research projects and drive new innovations. Larmon hopes Kiro will give users the confidence to take on projects they might not have attempted before.
How LMU gets the most out of Kiro
Larmon’s results come from how he uses the tool, not just having access to it. He now uses Kiro more than two hours per day and says it feels like he has “a three-to-10-person development team at my fingertips.” He offers this advice for getting started:
- Make time to explore: Carving out dedicated time to explore Kiro—and making it a priority for a couple of months—is key. The investment will likely pay off quickly.
- Pick a problem that needs to be solved: Choosing something that doesn’t usually get attention can demonstrate value early.
- Employ an iterative mindset: Getting the best out of an AI coding companion means engaging it actively, including asking clarifying questions, prompting it to explain its reasoning, requesting alternative approaches, and embracing a try-test-refine cycle. Larmon found that Kiro’s responses improved dramatically when he provided context on what worked and what didn’t, treating Kiro more like a team member than a code generator.
- Walk before running: Start small, put some guardrails in place, keep building, and then scale.
- Don’t believe everything at face value. AI makes mistakes and still needs a human in the loop. Checking Kiro’s outputs is essential. However, with continued iteration, teams will likely hit an inflection point, which is what Larmon describes as wondering, “Do I need to code anymore?”
- Make Kiro the most efficient developer possible: Developing a configuration project for Kiro—including a core set of steering documents, information about a specific MCP server, and compliance standards—pays dividends across projects. Anytime a new project starts, Kiro can “grab the information it needs from the config project.” This practice makes that knowledge portable to other Kiro users, who in turn become more efficient.
Looking ahead: Kiro workshops and campus-wide adoption
Larmon sees Kiro as transformational for LMU’s entire technical community. Later this spring, he’ll deliver full-day Kiro workshops and internal lunch-and-learns, aiming to have at least 20 people coding regularly with Kiro by summer. “The faster we can get teams on these tools, the faster we’ll see tremendous gains across any platform we manage. This is a real game-changer,” said Larmon.