Built Nimbleedge’s Federated Learning (FL) platform and individually designed, and managed their ML modelsampling, training and aggregation systems.• Critical in Architecture decisions on edgeML C++ SDK built on top of Onnx runtime. Which enabled us to distributewrappers for React-Native, Dart and Swift runtimes. To be deployed in iOS and Android applications.• Went live with a enterprise with 100M MAU. Systems I built were handling 20K requests per second at peak load… Show more Built Nimbleedge’s Federated Learning (FL) platform and individually designed, and managed their ML modelsampling, training and aggregation systems.• Critical in Architecture decisions on edgeML C++ SDK built on top of Onnx runtime. Which enabled us to distributewrappers for React-Native, Dart and Swift runtimes. To be deployed in iOS and Android applications.• Went live with a enterprise with 100M MAU. Systems I built were handling 20K requests per second at peak load. Wedid 50M inferences a day on EdgeML deployments.• Converted FL Research papers into Deployable and maintainable applications.• Implemented FL strategies such as FedAvg, Model Finetuning in the edge, Client Selection Polling, Secure Aggregation,etc.• Mentored and worked with Junior engineers and interns on deliverables and deployments. Show less

Part of the Governance team. Working on control plane services, Securing and governing privacy-vault.• Worked on RBAC and PBAC engines which enabled users to define granular access control, enabling users tocontrol access at table, column, row, location(based on IP lookup) and data-region level accesses by applications.• Designed and built service token based authentication for machine to machine Access management• Migrated from a single tenant architecture to a multi-tenant… Show more Part of the Governance team. Working on control plane services, Securing and governing privacy-vault.• Worked on RBAC and PBAC engines which enabled users to define granular access control, enabling users tocontrol access at table, column, row, location(based on IP lookup) and data-region level accesses by applications.• Designed and built service token based authentication for machine to machine Access management• Migrated from a single tenant architecture to a multi-tenant architecture. With data residency in multipleregions. Globally Rolling deployments with ¡3 minute downtime for enterprise users.• Worked on performance optimisations for the Data governance engine. Which reduced p99 response time from150 ms to 5ms at 15K rps by leveraging hierarchical caches and cache invalidation algorithms.• Migrated from an Arangodb based backend to Mongodb, with 80 Million records coming to about 1TB data, with zerodowntime To improve reliability and performance Show less

As a member of the platforms team, built Insurance audit platform organizing logs and information across 15 differentservices, built pdf digital signing, and claims payout and payin management services.• Improved developer tooling and compile-time over a rust codebase, by optimizing heavy dependencies and building infrafor remote code compilation to accelerate dev time.• During a period of management turmoil, stepped up and managed more than 150 releases over a period of… Show more As a member of the platforms team, built Insurance audit platform organizing logs and information across 15 differentservices, built pdf digital signing, and claims payout and payin management services.• Improved developer tooling and compile-time over a rust codebase, by optimizing heavy dependencies and building infrafor remote code compilation to accelerate dev time.• During a period of management turmoil, stepped up and managed more than 150 releases over a period of 7months coordinating across 8 different teams, besides regular development tasks.• pre-github-actions, Built an CICD platform with AWS device-farm to test, run and deliver test results on both frontendand backend systems. This reduced regression testing time by 80% Show less

Worked on internal tooling for software reliability, testing and deployment.Implemented on-demand end to end and integration Test running environments for developers with support for backend (linux) and frontend (MacOs native, Windows native, android, ios and Web) codebases.