Cultivating Career Fulfillment Through Impactful Work
Ganesh Kumar Gella, Director of Software Development for Amazon Bedrock generative AI services, describes how AWS’s culture enables his team to rapidly innovate and build with impact, and Deepak Nadig, Director and General Manager, AWS UX products and platform, provides an example of how teams across AWS are leveraging Bedrock generative AI services to transform customer experiences.
Please describe Amazon Bedrock Knowledge Bases and its impact.
Ganesh Kumar Gella: Bedrock Knowledge Bases is a fully managed capability that enables enterprises to harness the full potential of foundation models (FMs) by integrating their proprietary data and domain knowledge into an application. Bedrock Knowledge Bases is a powerful tool that helps businesses make the most of FMs by connecting them with their own company data and expertise.
Think of it this way: data is what makes AI solutions unique and valuable for each business, and Bedrock Knowledge Bases makes it easy to feed this data to AI models. By taking care of all the complex technical work behind the scenes, it allows developers to focus on building customized AI applications using their organization's best available information. This streamlined approach has made it one of Bedrock's most popular features.
What’s a key problem that Knowledge Bases is solving?
GKG: Amazon Bedrock Knowledge Bases provides a key differentiator for
customers looking to leverage FMs and generative AI. Rather than just fine-tuning an FM on a fixed dataset, Knowledge Bases enables customers to seamlessly integrate their own constantly evolving proprietary data sources, including both unstructured and structured data. This allows the FM to provide much more relevant and up-to-date information tailored to the customer's unique needs.
What were some key breakthroughs in developing Amazon Bedrock Knowledge Bases services?
GKG: One significant breakthrough was realizing we needed to build bridges
between our existing services and new generative AI capabilities. Another key realization was the need for additional services to address customer concerns.
This led to the development of Amazon Bedrock Guardrails, which allows
enterprises to put controls on what language models can say or not say. We also
created Amazon Bedrock Flows, enabling customers to design their conversation flows with a visual drag-and-drop builder for more predictability.
How are you enhancing existing services to meet future needs?
GKG: We're continually improving our services based on customer feedback and technological advancements.
We recognize that future enterprises will likely need multiple specialized agents working together, rather than relying on a single all-purpose agent. That's why we've already launched features like multi-agent collaboration in Amazon Bedrock Agents. By launching these advanced features now, we're enabling our customers to stay ahead of the curve and prepare for this next evolution in AI.
We've also enhanced Bedrock Knowledge Bases to not only answer questions from documents but also provide responses from structured databases, enabling natural language to SQL queries. And GraphRAG support allows for more comprehensive understanding across large documents.
Additionally, we're focusing on making our tools more accessible and user-friendly. Bedrock is available in Sagemaker Unified Studio, which makes it easier for enterprise employees to build knowledge bases and agents with a single sign-on experience. These enhancements are designed to transform how businesses interact with their data and customers, whether it's improving call center operations, boosting employee productivity, or enabling more sophisticated customer-facing bots.
How does AWS culture enable fast-paced innovation and pivoting?
GKG: At AWS, we understand that complete information isn't always available on day one, but that shouldn't prevent progress. Our approach is to make informed decisions with the available data while maintaining flexibility to adjust our course as we learn more through practical implementation. The ability to be vocally self-critical and pivot quickly is crucial to our innovation process.
Here's an example of what I mean: While building Amazon Bedrock Knowledge Bases, we realized the need to focus on information retrieval first. Due to this, we developed several features that enable improved retrieval accuracy, such as intelligently parsing documents using LLMs to better understand tables and images, contextual chunking, metadata filters, or automatic extraction of metadata, to name a few.
To give another example, we also realized that some enterprises have concerns about pushing their Generative AI applications to production, which led to us launching Bedrock Guardrails as an independent service, providing levers for enterprises to control the input and output from their applications. We then added contextual grounding as a feature to perform additional checks needed for Retrieval Augmented Generation (RAG) applications. Contextual grounding helps verify if an AI response stays true to the provided source material, essentially fact-checking the AI against the retrieved enterprise data. Most recently, we further enhanced this area with the launch of Automated Reasoning support. This feature helps with detecting hallucinations and provides verifiable proof that the model response is accurate.
DN: Our customers come to AWS to achieve efficiency. The AWS culture has enabled us to work back from customer needs, provide autonomy and ownership to highly talented teams to move quickly, and deliver customer experiences that enable customers to do their jobs efficiently.
Our products like the AWS Console, AWS User Notifications, AWS Console Mobile App, and Amazon Q Developer in chat channels (previously Chatbot) are built by many different teams. The AWS culture enables us to use customer data to align these teams while providing services and tools to automate their work and reduce undifferentiated heavy lifting. This has enabled us to partner effectively with user experience developers across AWS to deliver delightful customer experiences.
What skills and mindsets are you looking for in new hires for your team?
GKG: Curiosity and a willingness to experiment are essential qualities we seek in new hires. We want people who are not merely content with ideas but are eager to prototype and test new technologies. It's about identifying friction points in our daily work and applying innovative solutions to solve those problems.
The ability to deal with ambiguity is crucial, as many of the problems we are solving are first of their kind. It requires having the grit and perseverance to navigate uncharted territories and develop delightful products for our customers.
Closely monitoring whether a feature or launch is achieving its intended impact is equally important. Steering the product's next phases — based on real-world feedback and signals — is crucial for driving meaningful progress.
Any advice for your past self or for those early in their careers?
GKG: Approach disruption with an open mind and a positive outlook. Be willing to disrupt yourself and the status quo. We understand that new technological trends will continue to reshape what we've built, and by embracing these shifts with an open mindset, we can build better, future-proof solutions for our customers. Stay curious, adaptable, and willing to challenge your own assumptions and methods.
Additionally, always be open to taking on goals that may seem very daunting at first. With time and perseverance, you'll be amazed at what you can accomplish and the difference you can make, surpassing your initial hesitations.
