The automation tool is written in Typescript + React and uses a federated GraphQL service to communicate with the Java backend service. The service is deployed via Kubernetes and uses managed Amazon Aurora for storage. It was used to launch regulations in 40+ geographies affecting 100K+ listings in Q1'24

• Build and maintain the workflow system which is responsible for taking reviewed user inputs and performing a predefined set of tasks in order. Each task is designed to be reuseable, independent and idempotent. The task list is maintained as directed graph for each launch type which is resolved to its linear order using a topological sort at runtime
• Build and maintain support for additional task graphs to be mapped to a launch type to be run at different lifecycle stages of the proposal (eg: post-activation task graph for ops to monitor progress after activation reducing engineering involvement to validate launches)
• Build and maintain the review system required for launching a proposal. Added support for user roles, mandatory reviews by specific roles and an optional ability to self-review for smaller launches. For every section in the proposal, a list of mandatory roles is maintained in code. A review expection is created in the database when the proposal changes states and each row must have an accepted review for the proposal to be moved to activation.
• Continuously analyse past and predicted intakes to expand the scope of launches available via automation. Led the design and delivery of the roadmap for 2023 and 2024
• Simplified the Automation backend at Airbnb to reduce the time to launch new bundles. This was done to address the slow process of automating bundles due to high engineering costs. As a result, the backend effort for adding a new launch type was reduced by 80%.
• Decoupled user inputs from backend operations by leveraging an abstract factory module for bundle-specific implementation. This module translates user inputs into task inputs, streamlining the process of launching new bundles.
• Co-designed support for editability to allow Operations teams to self-serve updates to regulations
• Co-designed pre-launch analysis to compute the result of launches offline to enable safer updates. The system triggers an airflow job that computes the result of a launch before its activated to allow data driven decision making on the safety and correctness of a proposal
• Served as Primary POC for all automation support post transition to the Cities India team indirectly helping 100+ launches
• Onboarded 10 engineers to the automation tech stack, fortifying regulatory launches in the Cities India team. Established and led a weekly automation standup to ensure project timeliness, updates, and resolution of engineers' obstacles.
• Assisted in the integration of a new team members into the automation group, aiding in Netherlands and France subregion launches

• Translated legal and regulatory requirements into manageable product solutions autonomously. This aided Ops and Policy in their planning and launching processes.
• Led bi-weekly meetings with Cities Ops and Policy for prioritization, demos, and feature talks. This ensured the successful facilitation of over 25+ custom launches
• Provide consultation for complex regulatory launches (eg: NYC, New Orleans) including expanding system capabilities and guardrails for safer launches

• Designed and developed multi variate experiment support in the in-house experimentation product.
• Designed and developed a unified session solution to track user journey for accurate analysis of experiment effect on user behaviour.

We were building a product that would systemize the procurement process of goods at Udaan. We were purchasing and selling a lot of inventory but all the accounting was done outside the system. The goal was to build first class workflows around purchases, inwarding and reconciliation. This would help identify leaks and gaps in the financial process and prevent any potential for abuse.

The architecture would include a product master, location master, vendor master, the layer to raise purchase orders for products to vendors, the layer to match the goods received at the Warehouse against open POs and subsequently trigger the relevant financial workflow(s).

This was a massive overhaul of processes around warehousing, ordering, pricing and accounting. Key stakeholders across each domain were consulted and agreement was reached to roll out pan-india in phases over 6 months. The first warehouse was launched and the process stabilised in 1 month, 7 more were launched over the course of 2 months. Pricing and Ordering workflows were rolled out 100% across two categories with the rest to be ramped up in quick succession.

• Each location was mapped to kind of goods being bought/sold at location and had an approval matrix that would allow actions to be monitored and approved/rejected centrally
• Allow financial documents (po, credit notes) to be generated and sent directly to vendors via email
• Map purchase orders to vendor + location and only allow inwards of matching goods
• Every inward is checked against open POs for the vendor and only allowed if there is matching line item
• All deviations from the PO are recorded, staged in warehouse and require additional approval to be inwarded
• Every inward cycle to trigger the relevant financial workflows including vendor payments, adjustment notes, return notes etc

• Designed and developed a new architecture based on Domain Driven Design (DDD) for saving and presenting products data to users as part of migrating from a store based experience to a storeless experience. This included building a microservice that would communicate with multiple other services to process and present requested data to consumers.
• Designed and developed a Dynamic Salary System for honestbee deliverers (referred to as Delivery bees). The system takes into account various parameters for payout with support for peak periods and backdated changes.

• Implemented tier-based segregation for device types, enabling custom limits on simultaneous usage per device per customer, increasing device availability with minimal impact on genuine usage
• Designed and developed a diagnosis utility to notify users of test errors, reducing the load on the Tech Support team
• Created a rule engine to validate and map user-sent capabilities, providing extensive customization options for test suites
• Developed a Redis-based caching solution to prevent spamming from faulty user scripts and provide high availability for multiple services and async processing
• Developed custom client code on Windows/Android/iOS platforms to facilitate browser console logs, network logs, and native browser features for easy test debugging
• Designed and developed a daily summary service, sending users a digest of their tests based on their time zone, increasing dashboard visits by 20% and enabling faster problem notification
• Optimized and upgraded the REST API for easy modification and batch deletion of tests
• Maintained a real-time dashboard of customer tests with search and filter options
• Maintained a cross-region uploader service to collect and upload session logs scattered across multiple machines in multiple regions to S3

• Developed admin endpoints to empower the Sales team in managing user subscriptions independently, reducing reliance on engineering resources
• Created an internal newsletter tool for the Marketing team, enabling them to send targeted newsletters with time zone support to users
• Ensured smooth payment and order processing across multiple pricing plan iterations
• Conducted A/B testing using both external and in-house frameworks
• Led the complete redesign of the browserstack.com website, prioritizing page performance and responsiveness
• Oversaw the reskin and redesign of Product Dashboard(s)
• Implemented analytics and tracking utilizing internal and external data tracking services

• Automated the launch of new FC (Fulfillment Centre) for various services and tools pertaining to my team (Shipment Data Platform). Manual launching of new FC's was time consuming and prone to errors. Automation reduces human intervention and optimizes time.