The Ultimate Integration

View This Article in BOSS Magazine

Rackspace Technology is propelling the industry forward with game-changing innovations in Cloud, Data Analytics, and Artificial Intelligence

“Driven by a customer-first mindset.” That’s how Nirmal Ranganathan, VP and Distinguished Architect at Rackspace Technology, describes the trailblazing hybrid multicloud and AI/ML solutions provider. Rackspace delivers customized services and support for application, data, security, and AI solutions across private, hybrid, and hyperscaler public cloud platforms, spanning platform engineering, security, data engineering, analytics, application modernization, and AI.

Rackspace designs and builds scalable technology solutions and provides ongoing management and optimization so their customers can focus on building new revenue streams, increasing efficiency, and creating incredible customer experiences.

On the first anniversary of Foundry for AI by Rackspace (FAIR^TM) – human-centered solutions that seamlessly integrate AI into enterprises as a trusted employee and co-worker – we spoke with the 14-year Rackspace veteran to better understand Responsible AI, a set of principles for the design, development, deployment, and use of AI.

Nirmal Ranganathan

“Responsible AI is about making a positive impact through the use of technology for our company and our customers,” he began. Symbiotic is its first pillar. “It’s important for us to responsibly adopt AI, what we call ‘do no harm,’ align with the appropriate business need, provide proper safety accountability measures, and then create solutions that are transparent and explainable.

“Making tech consistent and predictable by eliminating biases, keeping it grounded on enterprise knowledge is one aspect of the second pillar, Sustainability. It also incorporates the environmental impact perspective,” he said. “These models require a lot of computing power to train, so making sure that the energy used isn’t driving additional environmental impact or that you're reducing that impact, through other means of providing a positive impact for the environment through the use of AI.”

Security, guaranteeing confidentiality of the models and the data behind them, and then preventing their misuse is an essential aspect of Rackspace’s focus, and is their third pillar.

Integrations pose daunting security risks and ameliorating them is part of Responsible AI. “There’s the traditional aspect of needing security at the network layers and application layers. Nothing changes there with respect to AI, so that's all intact,” he said. “Then there's security at the data layers. Beyond traditional data security measures to prevent breaches, it’s now essential to ensure data repudiation and guard data leakage through prompt injections and adversarial attacks.”

Safeguarding the validity or reputation of that data is crucial because compromised or unreliable data can lead to flawed models, inaccurate predictions and potentially significant business risks. Whether the data is used for training models or live inferencing, maintaining its integrity is vital. Rackspace implements a RAG model, a retrieval system designed to handle complex real-world queries over enterprise knowledge. “We send that data to the model and have it generate a response based on that context and apply guardrails and safety checks to prevent against potential injection attacks,” he explained.

While the data is safe from tampering, the model itself can have security risks. “If AI systems are starting to get interconnected and proper security controls are not applied, how do you ensure that the responses that a model provides are accurate, or that you don't have bad actors trying to manipulate the model into doing something that it wasn't intended to do?” he asked.

Guardrails and safety checks come into play, such as limiting how a model responds to certain inputs. “The challenge with these larger models and models getting even larger is that AI knows a little bit of everything, and it can be nudged, ‘prompted’ in a particular direction to respond in a particular way.”

Rackspace leverages generative AI to help their customers meet their most ambitious business goals. To start, they built two solutions. In August 2023 they debuted Intelligent Co-worker for the Enterprise (ICE^TM), a knowledge management solution that connects to all the enterprise sources of information such as SharePoint, Confluence, and other internal systems. The information generated by ICE^TM, a customized generative AI application based on enterprise knowledge, when combined with increased semantic and chat capabilities, allows Rackspace employees to create new insights, increase productivity and enable better customer outcomes.

The company’s longtime partnership with Amazon Web Services (AWS) is a boots-on-the-ground relationship, with AWS working hand-in-hand with Rackspace. “We were a launch partner for the AWS Generative AI competency, and we have also acquired the AWS Machine Learning and AWS Data and Analytics competencies in the past.” Ranganathan revealed. “We work very closely with AWS.” Together, AWS and Rackspace team up to explore solutions for Rackspace customers, helping them solve their most challenging IT and business problems.

Rackspace Intelligent Technology Assistant (RITA^TM), which launched this year, is an AI agent designed to boost user productivity by helping with daily tasks. RITA^TM can help answer questions and perform actions related to various IT services, such as setting up RSA tokens (multi-factor authentication applications that generate a random code at regular intervals), managing timesheets, and checking ticket status. “We used RITA^TM to automate our internal IT help desk ticketing,” Ranganathan said. “RITA^TM helped to streamline manual processes to just a few clicks.”

They are now working on AIDA^TM, a code generation tool that helps developer populations, leverage best practices for coding, consistency across teams and bringing older software up to date. As Ranganathan explained, “What are the gaps between an older version of Java or Python or a particular library and newer versions, and what changes do I need to make to the code? Typically, that takes a lot of manual effort.” These AI agents slice the time it takes to find these gaps more effectively and with higher quality and consistency.

A global accounts payable (AP) automation company, Basware sought out Rackspace to assist with automating code generation for their AnyERP product, a custom domain specific language (DSL) for integrating their client ERP solutions with their SaaS platform. The software integration typically would take as long as three months to implement – Rackspace reduced that time to about two weeks. Rackspace leveraged Amazon Bedrock to train a model to understand Basware's proprietary XML and developed code validation techniques creating a framework that enabled the company to deliver expertise at scale to all its customers.

“We've transformed technology over the last two decades. I've transformed in my career as well. I started out as a software developer and now I lead AI engineering teams,” Ranganathan mused. “Change is always constant, and we are focused on ensuring that as technologies and markets mature, we help our customers stay at the forefront of innovation.”