ODML (On Device Machine Learning) Tooling Team

Interned on the ML Metadata | TensorFlow Extended team- Evaluated the best cross-stack debugging tool for the team through a comparative analysis of metrics collected through a survey. Added support for internal users to the chosen tool for C++ backends, enabling many more teams. Enabled better observability by adding a log sink to forward all binary logs to the tool and Integrating the tool with the ML Metadata service.- Improved code health of the MLMD teams Go test codebase by… Show more Interned on the ML Metadata | TensorFlow Extended team- Evaluated the best cross-stack debugging tool for the team through a comparative analysis of metrics collected through a survey. Added support for internal users to the chosen tool for C++ backends, enabling many more teams. Enabled better observability by adding a log sink to forward all binary logs to the tool and Integrating the tool with the ML Metadata service.- Improved code health of the MLMD teams Go test codebase by refactoring and improving the readability of a few exceptionally long functions. Show less

- Primary owner of the Auctm web app. Responsible for architectural decisions, backend development, release planning, and coordinating with other stakeholders - frontend engineers, product managers, and designers.- Owner of sprint planning and task distribution, onboarding new engineers, and mentoring junior developers.- Actively took up interviews for hiring for senior developer roles.

- Primary owner of the Auctm web app. One of two backend developers involved in building the initial core set of APIs. Also the owner of architectural decisions, backend, and FE <> backend interactions.- Designed and built an insights generation engine based on effects and cause metrics shared by Data Scientists in the team. The engine automatically generates text representations and visualizations dynamically for any defined metrics.- The secondary owner of a team that built a… Show more - Primary owner of the Auctm web app. One of two backend developers involved in building the initial core set of APIs. Also the owner of architectural decisions, backend, and FE <> backend interactions.- Designed and built an insights generation engine based on effects and cause metrics shared by Data Scientists in the team. The engine automatically generates text representations and visualizations dynamically for any defined metrics.- The secondary owner of a team that built a dialler-based web app for faster and easier calling by Inside Sales Agents. Built the core front-end and part of the back-end to allow dynamic changes to different response types and validations to enable script flow.- Worked on DB optimization with PostgreSQL - reduced bloat by setting up more frequent auto vacuum on faster-growing tables, built partial indexes to reduce index size and query time on frequently queried tables. Show less

- Designed and built a flexible fraud detection engine, where frauds could be represented in the form of a PostgreSQL query. Basis the frequency of occurrence in this query different levels of actions could be automated on the fraudster.- Designed and built a queryable event log to track a customer's journey through the system.- Made two-way integrations between various CRMs and the SquadStack back-end.- Migrated the primary cache back-end from MemCache to Redis without any… Show more - Designed and built a flexible fraud detection engine, where frauds could be represented in the form of a PostgreSQL query. Basis the frequency of occurrence in this query different levels of actions could be automated on the fraudster.- Designed and built a queryable event log to track a customer's journey through the system.- Made two-way integrations between various CRMs and the SquadStack back-end.- Migrated the primary cache back-end from MemCache to Redis without any downtime.- Reduced the test suite run time by almost 10x by parallelizing it.- Made changes to the deployment process to allow cache invalidation of models with migrations to ensure zero downtime. Show less

- Worked on FastFox.com's backend, enabled various endpoints on the site using node.JS.- Part of the team that set up a producer-consumer architecture using the RabbitMQ broker to allow easier asynchronous process handling.- One of the two developers involved in designing and building a lead journey tracking system. This involved identifying and homogenizing data from different sources and storing it in a standardized, queryable form in MongoDB.

- Worked on the backend for MiniFastfox - an app interface for partner Brokers.- Was one of the three primary developers involved in building FastFox.com’s initial backend.- Picked up knowledge of node.js, PostgreSQL, Redis, RabbitMQ, Apache Solr, and MongoDB