Shruti Arora

Shruti Arora builds and explains workflows that connect cloud applications, browser interfaces and AI agents. Her work ranges from deploying web forms backed by private storage to connecting visual feedback with coding agents and checking their changes in the browser. In May 2026, she co-authored guidance on Amazon Nova Act’s HIPAA eligibility as a GenAI Specialist Solutions Architect at Amazon AGI, working with customers on production agentic systems.

Cloud applications and customer architecture

In 2022, Shruti was a Cloud Application Architect in AWS Professional Services’ Remote Consulting Team. With Kevin Rivera, Mark Carlson and Britney Tong, she co-authored a React and Formik application guide that connected browser forms to AWS infrastructure. Formik managed field state and validation; an API and Lambda function issued temporary upload and download links for files in a private S3 bucket. AWS CDK made the infrastructure repeatable to deploy. The example explained how the interface and its supporting services fit together as a working application.

Shruti also co-created a Lambda–Glue–S3 pattern with Harshitha Shashidhar. A Lambda function invokes a Glue job to convert CSV data to Parquet and save the result in S3, with permissions limited to the workflow’s needs.

By February 2024, Shruti was an AWS Solutions Architect supporting small and midsize betting and gaming businesses. She and Jake Hurst co-authored an account of Simplebet’s low-latency streaming architecture. For a service offering bets on individual plays, video delay affects whether viewers can follow the action in time. Their case study explained how Simplebet combined low-latency streaming with infrastructure serving users under different state restrictions.

Her May 2026 Nova Act article, co-authored with Abiola Babsalaam and Nishant Dhiman, addressed browser automation across healthcare portals, including scheduling, insurance verification and referral coordination. It distinguished service eligibility from customer implementation responsibilities: customers still need to configure access, encryption, logging and other required controls for their use of the service.

Connecting visual feedback to checked code changes

In Build with Perception Agents, presented with Emile Baizel, Shruti demonstrated how feedback on a rendered page can become an instruction to a coding agent and then a check of the resulting application. The shared workshop project connects two complementary tools:

• Visual annotation lets a person select an element, draw on the page or drop a pin, then export structured feedback as JSON. Element-specific feedback can include selectors and computed styles, so a request to make a heading smaller carries information about which heading to change and its existing properties.
• Rendered UI verification checks the application after editing. The shared implementation combines deterministic style checks with Nova Act executing user-flow scenarios, producing a report that makes failures available for inspection.

Shruti’s hands-on contribution was the integration workflow. Starting with a podcast landing page whose heading, colors and buttons needed changes, she connected a browser extension to a coding agent through a local bridge. Participants adjusted the extension’s permissions and added an Apply Changes button, turning exported feedback into a loop that edits code, runs verification and produces a report.

The workflow makes two decisions explicit: people define the desired design and behaviors, and verification checks whether the rendered application meets those expectations. Its value depends on both—the checks can expose a missed change, while the people using them remain responsible for choosing specifications and scenarios that capture what they want.