Pavan Chandrashekar

Software Engineer

Seattle, United States15 yrs 6 mos experience
Highly Stable

Key Highlights

  • Extensive experience in kernel programming and infrastructure engineering.
  • Proven track record in machine learning and anomaly detection.
  • Expertise in load balancing and performance optimization.
Stackforce AI infers this person is a Backend-heavy Infrastructure Engineer with expertise in kernel development and machine learning.

Contact

Skills

Core Skills

Load BalancingInfrastructure EngineeringMachine LearningData EngineeringKernel EngineeringEmbedded SystemsLegacy Systems

Other Skills

Load BalancerAWSAzureCapacity PlanningService DiscoveryDocumentationLog ProcessingFeature ExtractionSLURMData AnalysisData Ingest SystemUDPData ProcessingFlash StorageInfiniband

About

Generalist Major areas of experience: * Unix Kernel Engineer: Kernel Programming Experience on Solaris for about 6 years during which I added support for extended disk partitioning in OpenSolaris and also worked on key IO subsystems (device drivers) like storage and networking. Worked on the Infiniband stack of OpenSolaris, extensive performance engineering and usage of debugging infrastructures like DTrace, MDB on Solaris/OpenSolaris. Kernel Programming On Linux for about 5 years which included work on PCI-E based server-attached SSD storage. * Machine learning/infrastructure: Built a (log processing) pipeline for anomaly detection in logs employing machine learning. Designed a major part of the infrastructure including the pipeline backend, designed and implemented the feature extraction module/stage in the pipeline, mechanisms for debugging/analyzing the output/results of the anomaly detector, raw log/data analysis and runtime performance related work. Other work involved setting up and maintaining a SLURM cluster for running hyper-parameter tuning experiments and designing/implementing an extendable/automated attack system for automatic labeled data generation for the pipeline. * Building infrastructure for public ingress, external and internal load balancing at GitHub Goal is to solve problems via quality engineering and innovation.

Experience

15 yrs 6 mos
Total Experience
1 yr 11 mos
Average Tenure
--
Current Experience

Github

Staff Software Engineer

Jul 2019Present · 6 yrs 10 mos · Seattle, Washington, United States · Remote

  • I develop and maintain the GitHub Load Balancer, help improve the reliability, observability and performance of routing requests to different applications we load balance the requests to - both public facing and internal. There are many aspects to this role - capacity planning and management, service discovery, deployment complexities, a common design for a large variety of applications etc.
  • I've stood up AWS edge sites in the past to improve user experience, then migrated them to Azure. I've built key infrastructural parts of GitHub Enterprise Cloud with data residency - the site and load balancer design, the routing mechanisms, Certificate management to name some.
  • I've extensively documented different parts of GitHub Infrastructure and particularly wrote a comprehensive document bringing together all the pieces - which is now standard onboarding material for new hubbers.
Load BalancerAWSAzureCapacity PlanningService DiscoveryDocumentation+2

Oracle

Principal Software Engineer

Feb 2016Jul 2019 · 3 yrs 5 mos · Vancouver, Canada Area

  • Built a (log processing) pipeline for anomaly detection in logs employing machine learning. Designed a major part of the infrastructure including the pipeline backend, designed and implemented the feature extraction module/stage in the pipeline, mechanisms for debugging/analyzing the output/results of the anomaly detector, raw log/data analysis and runtime performance related work. Other work involved setting up and maintaining a SLURM cluster for running hyper-parameter tuning experiments and designing/implementing an extendable/automated attack system for automatic labeled data generation for the pipeline.
Machine LearningLog ProcessingFeature ExtractionSLURMData AnalysisInfrastructure Engineering

Raman research institute

Consultant

Dec 2014Apr 2015 · 4 mos · Bangalore, India

  • I built a data ingest system from custom hardware sending UDP packets over multiple (8) 1 Gbps links into a server. The data was processed in-line and a portion of it saved into secondary storage. I also implemented simple data processing using the AVX2 instruction set.
Data Ingest SystemUDPData ProcessingData Engineering

Virident systems

Technologist

Nov 2011Oct 2014 · 2 yrs 11 mos · Bangalore

  • Design and development of software for sharing and aggregating the flash storage across servers
  • using high speed interconnects - Infiniband and 10G Ethernet. The flash storage itself was PCI
  • Express based with very low IO latency. The challenge was to retain the low latency provided by
  • the base flash while adding layers of software onto it. Since the new layers introduced the network
  • into the IO path, Infiniband (RDMA) was used to minimize the additional latencies.
  • Also spent time on the caching solution (use of Flash Storage as a cache for disks) developing key
  • data structures required for a write-back cache. One was adoption of the FreeBSD RB-Tree
  • implementation into our code. The other was a partially sorted list that I implemented.
  • Performance enhancements and bug fixes in other components of the stack – primarily the Flash
  • Translation Layer (FTL).
Flash StorageInfiniband10G EthernetCaching SolutionsInfrastructure Engineering

Gluster

Principal Software Engineer

Jan 2011Oct 2011 · 9 mos

  • Responsible for overall performance and scalability of GlusterFS. Automated performance tests and
  • developed a framework to help measure and analyze systemic parameters during load across the
  • distributed setup.
Performance TestingGlusterFSAutomationInfrastructure Engineering

Sun microsystems

Member Technical Staff

Jan 2005Nov 2010 · 5 yrs 10 mos

  • Worked on different technologies while at Sun:
  • IPoIB administration remodeling
  • Design and implementation of a new IPoIB administrative model. This project involved changes to
  • the IPoIB driver, the Nexus driver, utilities and libraries. The change was to move away from nexus
  • based instantiation to administrator driven instantiation.
  • Solaris IPoIB Performance
  • Improvement of "IP over Infiniband" performance on Solaris. This involved an overhaul of the
  • existing solaris IPoIB driver to include features like Checksum Offload, Large Segmentation
  • Offload (LSO/TSO) and a re-architecture of locking to improve scalability. The end result was a
  • performance boost of 2-3 times the baseline, with greatly improved scalability.
  • Extended partition Support for Solaris
  • It initially started as a hobby project which later on gained acceptance into the mainstream Solaris code base. This project involved modifications to the following subsystems:
  • Solaris kernel modules – Utilities – Libraries - Solaris boot utilities - Solaris Devfs
  • Solaris Device Drivers Sustaining
  • Bug fixes to Solaris device drivers in various areas - Ethernet, USB, SCSI
  • During this period I developed insights into various subsystems of Solaris - understanding GRUB
  • and kernel initialization, device enumeration, etc. This also provided me with an opportunity to
  • learn extensive kernel crash dump analysis and usage of Dtrace.
IPoIBSolarisDevice DriversKernel Engineering

Motorola india electronics limited

Software Engineer

Apr 2004Dec 2004 · 8 mos

  • The project involved prototyping a PPP driver based on the Linux PPP driver to work on an RTAI
  • (Real Time Application Interface) based phone. RTAI is a flavour of RT Linux.
  • During this period I also did some work with a simulator to understand the Radio Resource Control
  • layer of the UMTS protocol.
PPP DriverRTAILinuxEmbedded Systems

Hewlett-packard

Software Engineer

Sep 2003Mar 2004 · 6 mos

  • Porting of a COBOL application running on a HP3000 mainframe to a HP9000 mainframe.
COBOLMainframeLegacy Systems

Raman research institute of astronomy and astrophysics

Project Assistant

Jul 2002Aug 2003 · 1 yr 1 mo

  • Device Driver work:
  • Developed a Linux PCI device driver for a couple data acquisition card manufactured by AMCC,
  • India and CDAC, India. Both drivers were for the 2.4 Linux Kernel.
  • FFT implementation using the MMX instuctions set:
  • Performance enhancement of FFTs (Fast Fourier Transforms) using the MMX instruction set. Used
  • the FFTW (Fastest Fourier Transform in the West - www.fftw.org) to generate C code and hand
  • converted it to assembly to get a speed up of 3.3 over FFTW.
Device DriversLinuxFFTKernel Engineering

Education

Bangalore Institute of Technology

BE — Computer Science and Engineering

Jan 1998Jan 2002

Stackforce found 100+ more professionals with Load Balancing & Infrastructure Engineering

Explore similar profiles based on matching skills and experience