Ankit Mangal

Developed and maintained dynamic ML platform team supporting over 200 models trained and deployed hourly, ensuring reliability and efficiency

• Designed and implemented Spark jobs to handle terabytes of data processing
• Engineered APIs powering dashboards for insights into model performance, feedback data, and pipeline health, leading to improved platform reliability and decision-making
• Redesigned Feature-Store (critical user path service providing real-time user and product features) to support QPS load exceeding 8 million and reduced client-side latency through cache enablement, facilitating onboarding of new clients like Ads and driving revenue growth( 3-4% increase in Ads revenue)
• Successfully scaled feature-store service to handle multiple Big Billion Day sales events
• Created Gradex, an internal request grading/rate-limiting library, enabling reliable scaling of various user path services such as Ads, recommendation, and search, particularly during peak periods like Big Billion Day
• Led platform and feature-store migration to Google cloud platform and reduced cloud platform cost (both storage and compute) by more than 70%
• Re-architected DB and system design of ML platform which allowed system to be extensible for new capabilities and added reliability

• Designed and developed an internal Resource and project management (RPM) portal which automated manual resource allocation task
• Developed frontend of the RPM portal using VueJs and Typescript
• Developed Backend for RPM application using Python and fastAPI
• Developed Integration of multiple Applications with the Splunk Phantom portal
• Worked on an health check and testing automation pipeline for Splunk Phantom Applications

• Developed a feature to create team combinations for a Resource and Project Management(RPM) Portal
• Actively contributed in discussions related to features of the RPM portal.

• Worked on a Network planning and nimble track tool used by Network planning officers.