Aug 2018 – Oct 2020 · 2 yrs 2 mos · Mumbai Area, India

• Member of the 3-member founding team in India office.
• Built ACT software from scratch. Filled in for ACT analytics, QA, dev-ops, DBA as and when
• required.
• Grew the India team from 3 to 15 currently. Also designed all the code challenges and
• interview process.
• Relevant technologies/platforms worked on – django-celery-stack, postgres, redis, SQL,
• docker, docker-swarm, terraform, metabase, presto-db, kafka, faust, snowflake, AWS.

Mar 2016 – Aug 2018 · 2 yrs 5 mos · Bengaluru Area, India

• Hyper-V expert at Nutanix.
• Helped build the current Hyper-V team in Nutanix.
• Worked closely with MSFT (after rigorous reverse-engineering) to support Windows 2016
• with Nutanix
• Worked on back porting kernel patches needed to support DDA (a new feature in Windows
• 2016) to Nutanix CVM kernel.
• Implemented the Nutanix SMI-S provider to support SCVMM 2016.
• Involved in on-calls related to Infra team and Hyper-V.
• Worked on LCM (Life Cycle Manager), responsible for life cycle management of all Nutanix
• software.

Aug 2015 – Mar 2016 · 7 mos · Hyderabad Area, India

• Worked on a new role in Windows Server 2016 - Network Controller. It is a highly available
• and scalable server role that automates the management, configuration, monitoring and
• trouble shooting of network infrastructure - both physical and virtual.
• Designed diagnostics, REST framework, disaster recovery and patching framework for
• Network Controller.
• Part of the framework/infra team for Network Controller.
• Involved in the entire product life cycle from planning to release.
• Worked on Logging, Diagnostics, Deployment, REST interface, Floating IP, Disaster Recovery,
• Update.
• Achieved Star performer of the team award - an annual award, at the end of first year.

Aug 2013 – Jul 2015 · 1 yr 11 mos · Hyderabad Area, India

Aug 2012 – May 2013 · 9 mos · Chennai Area, India

• The manual assignment of conference or journal submissions to reviewers can be a strenuous and complicated procedure. Apart from assigning the submissions to the experts that are scientifically competent to review them, the allocation is valid only if it satisfies a number of constraints. Considering the increasing numbers of both reviewers and submissions in current conferences, the whole process can be a difficult task for the program committee chair who is responsible for the allocation.
• The idea used here is to learn about the reviewers from their existing available papers and design of an automated system that can assist the program committee chair of a conference in the assignment of the submissions to reviewers. The system combines expert search with effective management of constraints. Initially, the system accepts as input the data for the reviewers, their publications till date, and the papers to be reviewed. Then, using expert search techniques the system identifies the reviewers that have appropriate knowledge of the topic(s) covered by the content of the paper. Finally, the papers are allotted to the reviewers, in the second step, taking care of the constraints and making the best assignment possible.

Jan 2012 – May 2012 · 4 mos · Chennai

• Worked on improving process scheduling in linux scheduler using a GPU.
• A set of Intel event counters was recognized and used as metric for the heuristic algorithm. The events were monitored for the current scheduled process in a rotation logic according to priority. The events were divided into two groups of high and low priorities. One event from each of the group was monitored in each iteration. Since, the ratio of high to low priority events was 3:10, high priority events were counted thrice as much as the low priority events.
• Designed and implemented a heuristic algorithm based on Ant Colony Optimization to be run on GPU to find a sub optimal schedule of processes. The event counters counted dynamically were used as metric for this algorithm.

May 2012 – Jul 2012 · 2 mos · Bangalore

• Worked on a research project on moving cellular baseband processing to cloud.
• Worked on estimating time scales over which cellular processing loads were predictable and designed algorithms to accurately estimate and predict these loads.
• Used these prediction algorithms to estimate the reduction in computing resource achievable in such cloud systems.