Snaplogic is an enterprise integration company (www.snaplogic.com).

• Implemented Snaplogic pipelines to integrate big data workflows.

• Evaluated the performance of Snaplogic Big Data pipelines against standalone Sparc scripts.

Backend engineer contributing to both the agent and server components for monitoring Big Data applications.

* ●  Implemented initial prototype of an agent for monitoring Pig applications.

* ●  Wrote sample Spark apps and monitored them using Driven's Spark agent.

* ●  Implemented REST APIs exposing the Driven ElasticSearch data model to the Javascript layer.

* ●  Implemented changes to a query parser to make enhancements to an internally developed query language against the ElasticSearch backend.

One of two engineers to build the Cohesity server backend from scratch.

• Wrote Cohesity manageability server backend using Golang and Martini/Gorilla frameworks.

• Designed various JSON data structures and REST APIs to expose to the Javascript layer.

• Implemented HTTP chunking (portions of the HTTP 1.1 spec) in the Golang server to support large file downloads from VMs stored in the Cohesity product.

• Implemented a proxy executable in C++ to allow the Golang server to interface with the remaining backend C++ components of the Cohesity product.

• Designed Google protobuf objects and implemented C++ functional and asynchronous programming under an event driver and component model.

• Worked with in-house implementations of components similar to Apache Zookeeper (for component registration and configuration) and an internal no-SQL implementation partially modeled on HBase.

Admitted to the Alchemist Accelerator, explored a few ideas over the course of a few months.

One of 2 engineers to create the .NET version of the AppDynamics product from scratch. Promoted from Senior to Principal Engineer after 1.5 years.

• Used bytecode instrumentation and sampling to collect timing, call stacks, SQL statements and errors for user actions.

• Implemented low-level C/C++ code for tracking code path execution and objects across threads.

• Reverse-engineered .NET frameworks to learn where to instrument code.

• Used Google protocol buffers for binary serialization to efficiently send data to central server.

• Implemented ConcurrentHashMap and other concurrent utilities in .NET.

• Lowered agent overhead to under 2-4% by profiling and avoiding primitive type boxing operations.

• Extensively refactored Java agent code to keep parallel structure to .NET agent.