An AI workflow can produce a convincing result while relying on incomplete or outdated information. In healthcare technology, that gap makes the engineering behind the model particularly important. Before information can support analysis or automated processing, teams need to know whether it arrived correctly, how it changed during processing, and which systems are permitted to use it.
Priya Vattikuti’s work addresses these underlying requirements. A Senior Cloud Data Engineer at Medtronic, with more than a decade of software experience, she works across Python, Spark, AWS infrastructure, and data processing. Her responsibilities connect the movement of information with the controls and monitoring needed to make it useful in an enterprise environment.
Reliable Pipelines Begin with Changing Data
Healthcare data platforms must accommodate information that changes over time. New fields can appear, existing structures can evolve, and downstream applications may require updates without repeatedly reprocessing an entire dataset. A pipeline that works against a fixed sample can become difficult to maintain when those conditions change.
Vattikuti’s experience includes Spark ETL frameworks supporting schema evolution and incremental loads. At Medtronic, she has built and optimized Python and Spark pipelines using AWS Glue for large scale processing and analytics. These responsibilities put the emphasis on how information is prepared and maintained, alongside the computing resources used to process it.
Schema handling and incremental processing address different parts of the same operational problem. The first accommodates changes in data structure; the second allows workflows to incorporate new or updated information. Together, they help establish a foundation on which analytics teams can build continuing processes rather than one time extracts.
Data Freshness Deserves Its Own Attention
A completed processing job does not, by itself, establish that a dataset is current. An upstream source may have stopped sending records, or an expected update may have arrived late. Teams therefore need visibility into the information moving through a pipeline as well as the availability of the infrastructure running it.
Vattikuti has developed serverless automation and monitoring with AWS Lambda, EventBridge, and CloudWatch to track pipeline health, data freshness, and operational service requirements. Her background also includes automated data quality checks. These measures help teams investigate whether a problem concerns job execution, missing inputs, or the condition of the resulting data.
That distinction matters when analytical applications depend on scheduled updates. Monitoring becomes part of the data delivery process, giving operators a basis for deciding when an output requires investigation before further use.
Access Controls Follow the Data
Within a shared platform, different services need different permissions. A transformation process may need to read source records and write prepared outputs, while an analytical application may require access only to selected results. Broad permissions can obscure those boundaries.
Vattikuti’s responsibilities include implementing least privilege IAM roles and policies for services including Glue, Lambda, Redshift, and Bedrock. Her work also covers healthcare and protected health information sensitive data. Using IAM roles and AWS Secrets Manager, she has implemented credential management that removes hard coded secrets from serverless applications.
These are concrete engineering decisions about which component can perform which action. They support enterprise security requirements without assuming that any single control establishes compliance or eliminates risk.
Connecting Data Engineering with AI
Vattikuti’s work extends into Amazon Bedrock integrations for classification, enrichment, and summarization, as well as SageMaker inference pipelines that return outputs to analytical data stores. This makes the relationship between data preparation and model execution especially visible. Model results must rejoin an operational data process if teams are to use them in analysis and reporting.
Her collaboration with analytics, machine learning, and platform teams spans ingestion through reporting. That breadth helps connect decisions made at the source with their consequences further along the workflow.
The practical lesson from this work is specific: healthcare AI depends on continuing attention to the data that reaches it. Pipeline monitoring, controlled access, and maintainable transformations give teams a stronger basis for assessing model outputs. Vattikuti’s experience places those responsibilities at the center of cloud data engineering as AI becomes part of enterprise workflows.
