Mar 2020 – Present · 6 yrs 1 mo

• Azure Bridge
• Bridging the features like Marketplace between AzureStack devices and Azure Cloud.
• Worked on K8s and other related open source tools for building containerized applications and deploying securely with monitoring & debugging set up.
• Azure Hybrid Network
• Developed services from inception to global availability:
• NFM: https://azure.microsoft.com/en-us/products/azure-network-function-manager
• Azure Network Function Manager offers a consistent management experience for pre-validated mobile network functions to enable a private LTE/5G solution. The service also enables a seamless cloud-to-edge experience for Azure Marketplace network functions such as SD-WAN.
• AOSM: https://azure.microsoft.com/en-us/products/operator-service-manager
• Azure Operator Service Manager transforms the operator service management experience into a modern cloud service. Simplify modeling with native Azure abstractions and ensure error-free operations using safe deployment practices. Consolidate multiple interfaces with support for Azure cloud and edge. Accelerate service deployments to take the lead in your market. E.g., Azure Nexus (https://azure.microsoft.com/en-us/products/operator-nexus)

Aug 2019 – Mar 2020 · 7 mos

• Substring Search
• Mentored on substring search evaluating various options and solving wildcard searches without impacting storage requirements and latenc impact.
• SLA Definition for Indexing Pipeline

Aug 2016 – Jul 2019 · 2 yrs 11 mos

• Code-Search for AzDevOps:
• Built reliable & performant Crawler: a crucial component for Code-Search which dictates overall load for pipeline, preemption, optimizations, etc. It crawls the VSTS version-controls to hydrate Search Engine.
• Built scalable crawlers both horizontally & vertically to improve the performance with reduced cogs.
• 11x faster crawling with horizontal scaling using distributed systems.
• 75% cost savings for worker-roles with vertical scaling. Making the service run on low config. nodes.
• Optimized multi-branch repo crawling.
• 20x faster crawling by
• pre-processing deDuplication files across branches to avoid re-processing content.
• shredded load of each branch to multi-threaded & distributed systems.
• Custom (SD) Indexing: allows users to leverage Search-Service for their code-base not hosted on AzDevOps.
• Architected SD-Indexers to efficiently host & index huge code-base (~40TB) as a single repo.
• Built performant multi-branch SD-indexing by pre-processing of duplicate files.
• 21x faster indexing with 25x reduced indexing load over Search-Service.
• Improved it to be more reliable, scalable & performant with reduced cogs
• 22x faster feeding & 12x faster code-base syncing by parallelizing feed & sync requests
• 440x faster repo-heal for huge repos (WDG), improvised in stages
• 55x by pre-processing multi-branch & isolating depots per repo to run in parallel.
• 8x by increasing intra-depot parallelism & using SSD disks. It also improved failed-items processing by 5x.
• 14x faster CI by isolating depots & repos to run in parallel.
• 57% Cost saving on VMs (due to above improvements)
• High availability & Disaster-Recovery(DR) for ES-Cluster: Its crucial to have DR plan for critical services to keep the service available/recoverable against disasters(E.g., earth-quake)
• Raising availability to 99.95% from 99.9% by re-Architected ES cluster for data-center level disasters.
• Architecting ES cluster for region level DR.

Nov 2014 – Aug 2016 · 1 yr 9 mos

• Code-Search for VSOnline:
• Cogs Reduction
• Saved ~30% cost of all resources in all service scale-units on production. Also, formulated scaling of service (for worker-roles as well Elastic-Search (ES) cluster [Search-Engine]) w.r.t. load required to cater. Hence, stopped reactionary changes, and enabled prediction of scaling of service up-front.
• Correlation in indexing Job/Operations
• Designed the missing correlation between jobs/operations running in independent context but for one end-to-end purpose. It helped analyzing the success/failures of operations end-to-end as well as much faster debugging of issues just from logs itself. It removed the requirement of production access for finding/root-causing issues related to indexing operations/jobs.
• Repository-healer
• Solved issues of search-results staleness and inconsistency of source code and indexed docs efficiently by utilizing least resources/operations of ES (Search-Engine) cluster as well as worker roles. This made sanitizing of indices (and hence search-results) possible must faster and less costly over resources as well as time.
• Account Migration (from an Index of ES)
• Solved down-time issues for service when a customer’s account was meant to move from an index. Achieved 100% availability for service even when the indices were meant to move from an index to another in pursuit of balancing load of indices, decommissioning an index etc.
• Monitoring and Alerting of ES-Cluster
• Set-up the basis for Monitoring and Alerts for ES Cluster, which was deployed on typical IaaS VMs, for the first time in organization. Till now monitoring and tools were available just for standard PaaS Services. This was a must requirement for proactive monitoring and finding potential issues without manual interventions.
• It removed the requirement of accessing prod search-cluster to find health/issues of the same. And helped in analyzing the perf and resource utilization and extrapolating the future needs to make service ready up-front.

Aug 2013 – Nov 2014 · 1 yr 3 mos

• I had worked in a couple of teams as SDET, mainly Azure Cache and VSOnline 'Search' Extension and spent a little time Microsoft WinSE 'beacon'.
• 1. Azure Cache, made sure Azure-Cache work flow is resilient towards failures like network outage etc.
• 2. Microsft WinSE 'Beacon', worked on an internal project called 'Beacon' which was used to create summary for a life of 'Bug' in the system, starting from customer end to the Engineer end where it gets triage and fix. Made sure this report is correct and it's processing is resilient towards race conditions etc.
• 3. VSOnline 'Search' Extension, it was a new project in VSOnline and I analyzed performance of different engines w.r.t. scale of data, to make a choice of 'The Engine' for this service. Also, for the chosen engine, analyzed the performance w.r.t. no. of tuning factors to achieve optimal configuration for the engine. Later worked on optimization to load this engine with high speed but still resilient towards failures.