AWS Showcases Pathways to Market With Generative AIAWS Showcases Pathways to Market With Generative AI

Q&A with Rahul Pathak, AWS’ vice president of generative AI and AI/ML go-to-market

Berenice Baker, Editor

October 17, 2024

4 Min Read
Rahul Pathak on stage at the AWS Generative AI and Data Day in London
Berenice Baker

Amazon Web Services (AWS) showcased how it supports companies to shift generative AI-driven technology from experimentation to production during its Generative AI and Data Day in London on Oct. 15. 

In this Q&A, AWS’ vice president of generative AI and AI/ML go-to-market, Rahul Pathak discussed how the company works with customers including Anthropic, BT and AstraZeneca to develop new services. He also talked about using generative AI to boost developer coding productivity and Amazon Bedrock, AWS’s managed generative AI service that offers safeguards and guardrails to ensure AI is used responsibly.

AI Business: How does AWS work with customers to take AI-powered solutions into production?

Rahul Pathak: It starts with what the customer needs and then working backward. That’s been a core part of our DNA, customer obsession, but the real success factor is figuring out what business value you're trying to drive. Once you have that, you want leadership buy-in and then we figure out what we have to offer that can best help customers achieve that objective. The key to AI solutions being successful in production is having that business outcome.

We’ve seen AI supporting sports events including the Olympics and Wimbledon. One of the customers here today is F1. What did AWS work on with them?

Related:Amazon Web Services Earmarks $10B Investment for UK Data Centers

F1 is data-driven, they've got telemetry about every second, every millisecond of what's going on. It's been a phenomenal partnership covering everything from making F1 media assets available in the cloud and having generative AI as a tool for broadcasters to interact with that, to new ways to interact and engage with the audience.

We've helped with vehicle telemetry and the 2022 car design. It’s about taking all of the F1 data and historical footage and then using AI, machine learning and generative AI techniques to index that. This helps with predictions and creates a more engaging experience for F1 teams, broadcasters and ultimately the fans.

How do AWS tools help companies build their own code?

BT has been using Amazon Q Developer to drive their internal developer productivity. Amazon Q Developer is a generative AI assistant for all phases of the software development lifecycle, from planning a project to creating code, to deploying and testing it and then to upgrading it over its life cycle.

BT has really leaned into generative AI and Amazon Q Developer specifically and they're able to get about 300,000 lines of code auto-generated with a 33% acceptance rate. So that's 100,000 lines of code that AI is written on their behalf with oversight that's been deployed and is creating value. That really improves developer experience, improves productivity, reduces the amount of drudgery in the day-to-day and ultimately allows them to do more with the teams they have.

Related:AI Risk Mitigation: The Role of Testing

How does Amazon Bedrock, AWS’s generative AI app-building platform, ensure responsible AI?

Safety is a huge part of putting AI to work responsibly and it's been a big area of investment for us. The models available in Amazon Bedrock all have some protections built in, but on top of that, Amazon Bedrock does a lot of automated abuse prevention, like detecting unsafe words, looking at inappropriate things and trying to automate safety on top of what the models provide.

We've seen up to an 85% improvement in reducing hallucinations and staying focused. We also have a capability called Amazon Guardrails, so customers can define their own policies to sit on top of the automated process to give it contextual grounding. It’s all aimed at making it easy and safe for customers to put AI to work.

What are some of the biggest generative AI challenges your customers come to you with?

Customers are looking for how they can get value, drive a business outcome and get that right. Return on investment starts with defining what you are trying to accomplish and then we can work back from that to say, how can AI help? The goal is not to use AI for AI's sake, it’s to drive a differentiated business outcome.

Data is our customers' key differentiator. As we democratize model access, customers will differentiate by bringing their unique data privately and safely, combining it with those models and building interesting apps. Helping them connect those dots is where we've been focused again.

AWS offers a large range of AI tools. How do you help customers select what’s right for them?

We do it through being transparent. There is a model evaluation capability in Amazon Bedrock, we've got model cards where we walk through the model capabilities and then we work closely with our customers to match the model to the use case.

We often find that customers are using more than one model and these models also evolve with customer use cases. Every couple of weeks there'll be a new announcement about an advancement in models and customers don't want to have to rewrite their apps every time that happens.

That’s another piece, giving them a stable foundation that's still flexible enough to let them be future-proof and take advantage of the innovation that's still to come. I don't think that pace is going to slow down anytime soon.

About the Author

Berenice Baker

Editor, Enter Quantum

Berenice is the editor of Enter Quantum and co-editor of AI Business. Berenice has a background in IT and 20 years of experience as a technology journalist.

Sign Up for the Newsletter
The most up-to-date AI news and insights delivered right to your inbox!

You May Also Like