Our Karsun Innovation Center (KIC) Interns explore complex modernization problems, experiment with emerging technology such as artificial intelligence (AI), and engage with experts on our research and development team. Every summer, we feature our interns’ work on the Karsun website. In this interview, Computer Science student Samhita Kumar shares her experience with the AI code pairing tools included in Karsun’s ReDuX AI, front-end design using component libraries, and task management automation.
Finding Her Next: Software Development, Cryptography and AI Innovation
First, please tell us about yourself. Where are you going to school? What are you studying? What do you like to do in your free time?
Samhita: I’m Samhita, a rising sophomore at Yale University studying computer science. In my free time, I enjoy reading, visiting museums, and baking. I’m also part of my school’s moot court team and love learning about Constitutional law.
What do you want to do after this internship? What are your career goals?
Samhita: I hope to build on the skills I’ve acquired while here, and to apply my knowledge to real-world problems. I aim to pursue opportunities in software development and cryptography, and to continue working with new innovations like ReDuX.
Full Stack Development and Generative AI
Karsun’s ReDuX AI uses generative AI to make code, data, and access control recommendations. In addition to their work exploring the use of AI for modern software development, we also challenged our interns to identify and propose enhancements to our AI toolkits. Samhita applied her full stack development skills to this challenge.
Could you share a little bit about the project you worked on as part of this internship? What challenge does it solve? What technologies and tools are you using?
Samhita: I have primarily been working with KIC’s ReDuX AI code pair tool, AppPilot. I used Java SpringBoot, PostgreSQL, Podman, Testcontainers, Nx, React, and Jest to build a full-stack project while testing AppPilot. My front-end design pulled in several component libraries such as Metrostar Comet. I also created a Task Manager project with CRUD functionality and user accounts to strengthen my understanding of the technology stack. Finally, I helped find and resolve defects within AppPilot.
Flexibility, Adaptability, and Innovation
What is your favorite part about working with the Karsun Innovation Center? Is there a weekly meeting or ritual you enjoy? The opportunity to learn more or get a new certification?
During our biweekly “Show, Don’t Tell” meetings, I’ve been able to learn about the projects that the Karsun Innovation Center is working on. It has been amazing to explore so many different technologies and to see the different approaches that can be taken to address blockers.
What is your biggest takeaway from your experience as an intern at Karsun?
It is definitely the importance of flexibility. I’ve gained so much practical experience and observed how unexpected issues can arise at any moment. To tackle these challenges, I learned to remain adaptable and prepared for anything. This approach has proven incredibly helpful, and it’s one I will definitely carry forward in my career.
Samhita worked alongside Karsun Innovation Center experts throughout her internship program. Discover how Karsun experts are modernizing for every next in our Innovation Center by accelerating transformation with the ReDuX AI toolkit.
Recording live from the AWS Summit Washington, D.C., Karsun’s Badri Sriraman sits down with John Gilroy and the Federal Tech Podcast to discuss all things AI, accelerating legacy systems modernization and hyperlocal contextualization. Badri is the Vice President of the Karsun Innovation Center. Here, teams work to eliminate or reduce friction for the agencies modernizing their legacy systems. As part of that work, they developed ReDuX AI, a toolkit using AI resources to address common issues related to modernizing these older, more complex legacy systems. Throughout the interview, Badri and John discuss the costs and security risks associated with staying on current systems and the opportunity for AI to provide insight through hyperlocal contextualization to tackle these challenges.
Addressing the O&M Problem
The interview begins by acknowledging the drag aging infrastructure has on these agencies’ enterprise systems. The issue becomes intractable when operations and maintenance (O&M) costs become so high they take up the budget that would otherwise be used for modernization and systems enhancements. A 2023 Government Accountability Office (GAO) report examined the 10 critical systems it identified most in need of modernization. Some of those systems were over 50 years old and, in total, cost the government upwards of $337 million annually to operate and maintain.
In addition to the O&M costs, complex relationships between different systems components, older programming languages like COBOL, and outdated documentation all contribute to the difficulty of modernizing mission-critical legacy systems like those studied. As Badri reveals in the interview, artificial intelligence (AI) can be used to address many of those concerns. Moreover, those same AI tools may accelerate the transition, further reducing O&M costs as agencies move away from these legacy systems.
Eating the Elephant
In the interview, Badri shares a key component of Karsun’s AI-accelerated modernization methodology, hyperlocal contextualization. In a traditional modernization project, teams typically take an incremental approach. To “eat the elephant,” teams go one step at a time, optimizing as they modernize to reduce costs. The analysis required to peel away each part of the legacy system could slow down modernization to a yearslong process in highly complex systems. Throughout the process, these teams must untangle how different parts of the system integrate and work together.
Alternatively, using AI teams still move incrementally, avoiding the pitfalls of Big Bang Modernization. However, they also use AI to identify and map these relationships within legacy systems. This AI assistance helps those teams move at a rapid pace. Working together with the human team and stakeholders, the AI can quickly create a fuller understanding of the system, its impact on the mission and provide new insights into optimization opportunities during the modernization process. This analysis is the first step in using AI for hyperlocal contextualization.
The Security Imperative
An equally pressing concern is rising security threats. The 2023 GAO report found among the chief concerns for these aging systems were outdated hardware and security vulnerabilities. The same processes that help modernization teams optimize to reduce O&M costs can also be used to move from less secure systems to more secure systems based on industry standards and best practices.
Jumping in with Hyperlocal Contextualization
With more advanced large language models (LLMs), AI tools cannot only write the code but also review, debug, and make recommendations. Powered by AWS Bedrock, ReDuX AI not only takes the insights from its eat the elephant analysis but also makes recommendations to Karsun developers using the context generated from those insights. Plus it can then incorporate best practices and security policies into those recommendations. Using a tool like ReDuX AI, the security optimization recommendations are customized to the complexities found in that legacy system, resulting in a modernization process designed to optimize costs and security. Badri discusses this hyperlocal contextualization in greater depth in the interview.
The future is bright. In a 2019 report, the GAO shared agencies provided 94 examples of successful modernization initiatives over the five years studied. While some of the most challenging projects remain, AI tools, like those used by ReDuX, create a future where the remaining mission-critical systems modernize a possibility. Check out the full interview at www.theoakmontgroupllc.com/ep-161-how-to-overcome-the-challenge-of-legacy-systems/ or visit GoRedux.AI to learn more.
Kartik Mecheri joins GovLoop’s Featured Contributor program to share his insights on artificial intelligence (AI), machine learning (ML), modern software development, architecting for cloud-native environments, data solutions, and enterprise modernization for the public sector. The co-founder and Chief Architect of Karsun Solutions, Kartik’s achievements include spearheading large-scale digital transformation projects. His expertise ensures that Karsun delivers cutting-edge services to its customers while enhancing their customers’ internal processes. He has earned industry recognition for excellence in technology solutions, keeping Karsun at the forefront of technological advancements and customer satisfaction.
GovLoop’s core mission is to help government employees to do their jobs better. It provides free resources, blogs, online training, in-person training, and online courses on topics relevant to anyone working in public service. This new series is an extension of Kartik’s position as a trusted advisor to senior government executives on the latest technologies.
Throughout his featured contributor series, Kartik will combine Karsun’s experience modernizing complex legacy systems for federal government agencies with emerging technology research from by the Karsun Innovation Center. In addition to its research and development unit, the center also builds resource toolkits to assist Karsun’s modernization teams. As part of his column, Kartik will share Karsun’s experience using these tools to accelerate modernization.
This includes sharing Karsun’s experience implementing its ReDuX AI toolkit. This toolkit uses AI to address the challenges associated with migrating complex legacy systems. Its AI tools analyze legacy systems built on older coding languages, such as COBOL. Next, it produces visualizations and other information on the structure of the legacy system. With this enhanced insight into the complex relationship between system components, teams can effectively plan incremental modernization of the system without disrupting the current mission-critical parts of the application. Next, the toolkit resources use the information from the system analysis to make recommendations, enabling teams to generate code more efficiently and securely than code generated by human teams without access to those resources. More information on ReDuX is available at GoRedux.AI.
In his first post examining AI-assisted modernization, Kartik delves into the challenges and opportunities presented by the use of AI for code generation. Here, he discusses the evolving software development space as teams use AI to analyze codebases, security and regulatory compliance, and increased efficiency for mundane development tasks. For more from Kartik, follow his Featured Contributor series at https://www.govloop.com/author/kmecheri/.