Jul 2024 – Present · 1 yr 10 mos

• Led the development of BharatMLStack a end-to-end ML Platform, as a committer and maintainer, enhancing data accessibility and economics of AI/ML at Meesho. Leading open source initiative for the same.
• Spearheaded the Machine Learning platform at Meesho, driving innovation and optimization in ML algorithms and processes.
• Architecting the overall demand organization architecture, optimizing operational efficiency and scalability to match Meesho's growth.

Jan 2022 – Jul 2024 · 2 yrs 6 mos

• 1. Planned and owned road maps for migrating batch mode inferences to near real-time and real-time setup. The migration helped in reduction of over computation cost by xx%
• 2. Designed and developed in house online feature store for super-fast feature retrieval during online inference. Creates and manages multiple pipelines of the store, each pipeline handling events in order of billions per day, and allow flexible storage selection for each tenant, with visualisations and other aspects.
• 3. Managing a highly motivated team thats capable of delivering massive migrations to real-time personalisation in weeks.
• 4. Designed the experimentation framework for real-time personalisation with systemic controls on isolated audiences and overlapping audiences, and capability to version the configs used for experimentation for analytical purpose.
• 5. Designing and developing intra app-session personalisation capabilities to make personalisation more effective and fast.
• 6. Lead the team in designing cloud agnostic infra structure for model deployments across clouds
• 7. Lead the team in designing vector similarity search platform
• 8. Lead the team in designing and developing framework for computing higher order interaction that led to server cost reduction
• 9. Lead an inner sourced team to design and develop mCache, that is Meesho’s own serverless caching, reducing developer involvement during scaleing activities and cost spent on caching vendors

Apr 2021 – Jan 2022 · 9 mos

• 1. Designed and owned the 1 year plan for discovery in general, feed in specific to migrate out of monolith. Able to deliver 95% of the plan by the end of the year along with numerous other product projects.
• 2. Planned and executed migration for widget infra to MongoDB based solution.
• 3. Designed and developed fallback feed service for powering listing pages, handling throughput in order of x10k Qps with p99 latencies of ~15ms

Oct 2019 – Mar 2021 · 1 yr 5 mos

• Search and Feed ecosystem
• Ads and monetization
• Aggregation layer management
• Migration multitenant system
• API Gateway
• Product features

Feb 2019 – Sep 2019 · 7 mos

• Working to scale the micro-services to handle ever increasing hit rates.
• Developed public facing aggregation layer from scratch using Hystrix and manifesting various data intensive architectures for scalability and resiliency of all downstream systems.
• Supplier scaling through marketplace APIs
• Product features