Vikas Kumar Email & Phone Number

Chicago, Illinois, United States

Developed, optimized, and maintained a Tempus gateway service that provides a single entry point for multiple data modalities (hl7, fhir, imaging) to providers and third-party labs that require ordering workflows. Implemented OAuth 2.0 Client Credentials Flow for authentication and authorization, WAF rules, rate limiting for a publicly exposed service.Introduced FHIR (Fast Healthcare Interoperability Resources) as a healthcare data format for accepting clinical sequencing orders through EMR (Electronic Medical Record) systems for 150+ institutions. With this, Tempus established a footprint in adopting FHIR improving upon the rigidity of HL7v2 as a data exchange medium.Introduced system-level functional and non-functional (performance) automated testing in tightly-coupled event-driven microservices that vastly helped in reducing the number of observed major incidents and established strict SLAs for publicly exposed services.

Pioneered developing an analytics platform using event sourcing (∼3M events per day) with near real-time dashboards for observability for the orders received by Tempus through EMR systems and on sent-out results. The BI reports were used to track the lifecycle of each order received from 1000+ healthcare institutions along with completion checks and failure alerts for each order/result.Designed, developed, optimized, and maintained batch processing pipelines (millions of files per batch) that adhere to ELT principles, i.e., consume, process, and store unstructured/semi-structured ECG/Dicom/Molecular data received from third-party integrations.Designed a rate-limiting solution in a distributed environment using a key/value database as the counter to avoid overloading Tempus services with HTTP calls. Built real-time dashboards to observe the workload/throttling received within the system.

State College, Pennsylvania Area

Funded by National Energy Technological Laboratory, developed a physics-informed deep learning method that combined numerical simulations and machine learning to explore CO2 storage site behavior in various operational scenarios. Physics-based partial differential equations are used to modify the loss term leading to 5-10 times better performance than data-driven models.

Master Of Science - Ms, Computer Science, 3.96/4

Courses: Deep Learning, Numerical Linear Algebra, Distributed Systems, Computer Vision, Artificial Emotion Intelligence, Machine Learning.

Bachelor Of Technology - Btech, Computer Science, 9.2/10