Aug 2024 – Apr 2025 · 8 mos

Apr 2022 – Aug 2024 · 2 yrs 4 mos

• Developing rider hub
• My team is making a new platform for rider onboarding with a vision to convert qualified applicants to riders within 24 hours. This new rider hub is made using microservices deploying technologies like Ruby on Rails, NextJs, and Terraform to make eventbridge, appflow, gateways, dynamoDB, S3, and other resources. In this project I have made the login and authentication system for riders and rider applicants, a rider comms system to send them notifications about the progress of their application, created Infrastructure using code, enhancements to send applicant data to analytics pipelines, and created a few UI screens for the hub.
• Working on Rider onboarding and internal toolings
• In this project, I worked on creating an apply page for rider applicants which serves as a gateway to created rider application pipeline to convert an applicant to rider, where they complete other required checks and stages to become a rider. Here I have also worked on tools used by internal teams to manage onboarding of riders, resources & training related tools and manage those riders.

Jul 2021 – Apr 2022 · 9 mos

• Horizontal Pod Autoscaler: The load on our application was skewed concerning time. For different clients, the server had the most load for a certain time during the day. And to manage the highest load of the day without crashing, servers had to run at max load 24X7. Kubernetes gives the flexibility to scale horizontally very fast. I published load-defining metrics to datadog and based on those metrics Kubernetes will scale applications whenever there is a high load. At high load time, the application runs with more number replicas and at all other times, it runs a bare minimal infrastructure. This has reduced deployment costs by 30-50% in multiple clients.
• Intelligent file scheduling- To process a file, the application needs to resolve client and security identifiers for each record. To identify this data it needs to fetch data to fit in rules defined from multiple upstream applications. For multi-client files, this processing run in a shared resource, but it fetches data from client pods to resolve the data. All clients and upstream applications have different latency. I designed and implemented an algorithm to start the process intelligently by checking the availability of upstream resources to reduce queued time and use resources more efficiently.

Jun 2019 – Jul 2021 · 2 yrs 1 mo

• SLA-based Scheduling algorithm: File arriving into platform first passes through parsers in FIFO order for data transformation and security checks. Sometimes when files are delayed, other dependent processes are used to wait for the file. It needed manual intervention to prioritize such delayed files. I designed the algorithm to prioritize files based on their SLA, arrival time, and historical processing time. Now if any file is delayed, it informs the stakeholders and when the delayed file arrives, it gets processed in its relative order of importance.
• Making system distributed: Earlier apps were running on a single machine, which had a lot of limitations including hampered availability, high deployment times, and cost. I made applications compatible to run on multiple servers by integrating hazelcast for distributed caching, ActiveMQ for messaging, zookeeper for distributed locking. I made 3 applications distributed, now these apps are highly available and AWS cost is 20% less than what it was in a single server. While reducing the cost for deployment, this process has increased the throughput and made applications highly available.
• Hosted on-prem Hazelcast: cloud-based Redis/Hazelcast lacks authentication and hence using a shared distributed cache for multiple services was not secure. And individual cloud instances of each service were stacking up the costs. I deployed hazelcast in embedded mode as a sidecar to the applications.
• Database query optimization: While making applications more available and distributed, load on the database has increased. I have optimized a lot of frequent and long-running queries. I have explored table partitioning, index hints, optimal indexes, etc to keep database resources in check. I was able to bring down database load by optimizing queries and I was able to get more throughput for application by using the same database resource.

Jun 2018 – Aug 2018 · 2 mos · Hyderabad, Telangana, India

• I have worked on Business Intelligence Application along with SQL Server. I have developed an application to automate an internal data verification process to make deployment easier and faster. The Application can be used to verify the data across OLEDB database, cube and Marts on different servers. This Auto Validation Tool gets triggered by SQL Server job agent, performs validation and sends verification email.
• Technologies Worked on: Visual Basic, Power BI, SQL

May 2017 – May 2018 · 1 yr · Mandi, Himachal Pradesh, India

• At IIT Mandi Catalyst, a Technology Business Incubator by startup India, I played an active role in planning and managing startup summit events like Himalayan Startup Trek. These events provided me with an opportunity to interact with budding startups and industry experts from a wide range of domain. Having worked within the Catalyst for over 1 year, I have developed a wide range of soft skills.
• Key Responsibilities: Web Development, Graphics Designing, Digital Media Campaigning, Event planning and management

Oct 2016 – May 2017 · 7 mos · Mandi, Himachal Pradesh, India

Oct 2016 – May 2017 · 7 mos · Mandi, Himachal Pradesh, India