Jul 2021 – Oct 2022 · 1 yr 3 mos · Gurugram, Haryana, India · On-site

• ‣ Led the initiative for Logistics Cost & Delivery Time Reduction.
• Created feature extraction & reranking module bringing better control over customer’s feed. (handling ∼100M requests each day)
• Performed extensive A/B experimentation on derived feature sets to arrive at an algorithm for generating optimal restaurant feed.
• This led to a monthly savings of ∼$400K in logistics cost and ∼3 mins reduction in delivery times to customers.
• ‣ Built and launched a new line of business - “Zomato Grocery” - Led the “Search & Discovery” Team for the new LOB.
• Store Listings: Prepared the design and a neat, scalable implementation for the store listings on home page, targeting a higher customer engagement, GMV and OTR.
• Product Search: Designed and built autosuggest for powering stores, products, categories search from scratch. Introducing features like type-ahead search, capturing customer intent etc.

Jun 2019 – Jun 2021 · 2 yrs · Gurugram, Haryana, India · On-site

• ‣ Built and launched a new line of business - “Zomato Nutrition”
• Worked on building 2 of it’s core microservices - Order Management Service & Content Management Service.
• Built the Event Pipeline to process order events, maintain states and data consistency in our event-driven system.
• Built the entire Post-Order Experience System single-handedly.
• Built the entire Insights & Alerting System utilising StatsD, Prometheus, Grafana, NewRelic etc. for providing real time insights and alerting in case of anomalies.
• ‣ Reviews 2.0 and Turing Engine - A complete revamp of the “Zomato Reviews” platform from a user perspective, introducing the concept of “tags” to organize the plethora of textual content into crisp, canonical structures. This form of UGC ingested was in turn used to power features like “Reviews Search” and “Review Highlights”on app. Utilized various NLP techniques during the process to build an ML model for “Aspect-based Sentiment Analysis”.
• ‣ Zomato multi-rating system - Worked on migrating Ratings platform to a new multi-rating system. Carving it out from monolith to an independent module.
• ‣ Feedback Service - Worked on creating a domain-driven design oriented, multi-tenant microservice that is now being used across all business verticals for capturing customer feedback. Inspired by Hexagonal architecture, we made this an easily pluggable and evolvable system. It has the highest avg. throughput across all services. (400k rpm)
• ‣ Reviews Service - Worked on migrating “Reviews Platform” from monolith to a microservice, rewriting the entire codebase from scratch keeping scalability as the main focus.

Dec 2018 – Feb 2019 · 2 mos · Gurugram, Haryana, India · On-site

• Utilized Bag-of-Words, TF-IDF and Word2Vec for feature extraction from a dataset of emails ranging into thousands to empower sentiment analysis and predict the grammatical flow of emails. Furthermore, created and trained an ML model (SVM) utilizing said features, achieving the final objective of grading emails with accuracy comparable to human graders.
• Utilized ULMFiT and fastai library to enable transfer learning for text classification.