Ritvik Joshi

Matching and Reconciliation system (MARS)

• Contributed along with team of 3 to re-architect and rewrite the system to scale from 10L documents to 50L+ documents supported in the system (5x). Also, the time to matching was reduced for large reconciliations from 12hr+ to 1hr (80% faster), leading to an increased ARR by 15x from 6Cr (2022-23) to 90Cr+ (2023-24)
• Conceptualized and implemented the "Near-Exact" in-house matching algorithm for invoice number similarity using historical data. Defined and streamlined algorithmic rules, reducing the number of rules from 50-2000 to less than 20, and eliminated false positives, significantly improving accuracy (0% FP) and efficiency in invoice processing.
• Designed and implemented an ingestion system to automatically take user actions from an excel sheet which reduced customer's TAT to take ITC actions during filing process from 2 hrs+ to less than 5 mins
• Addressed high network bandwidth usage and reduced downstream api calls by caching the calls using redis and local cache. This led to 5x improvement in p80 & p95 for ITC action processing time
• Independently owned multiple features end-to-end, identifying requirements, solutioning, HLD and LLD,coordinating with teams, and setting up performance metrics and alarms.
• Mentored junior engineers as well as interns, providing guidance on best practices, code reviews, and professional development.

Clear Automation Framework

• Independently owned & built a generic framework for SDETs to write end to end test cases for any microservice using karate's data driven model
• Reduced shipment time from QA to PROD from 7 days to less than 1 day

• Designed & developed GET/POST/PUT REST endpoints for 20+ ITSM workflows
• Applied SOLID principles in design phase to keep application scalable & maintainable
• Used in-memory cache to store meta information (user profiles, groups, org-paths etc.) & improve response time by more than 5x
• Implemented session-based authentication mechanism which helped to automatically expire sessions after certain interval
• Automated cleanup of inactive API users using python script, based on last access activity and thus completely removing manual effort
• Automated user, org-path & group updates for company reorganization,reducing process implementation time by 80%