Aug 2024 – Present · 1 yr 10 mos · Hybrid

• ■ Created and aligned elaborated long term strategy for User Generated Content of Flipkart.
• ■ Ratings and reviews always have a negative bias. Customers who had positive experience barely feel motivated to provide feedback. Solving this by designing smart incentives and nudges while balancing cost vs impact and ensuring unbiasedness. Impact: increased response rate by XX%, topline by XXXCr.
• ■ For the products with too many reviews, users feel overwhelmed to read all. Solving this problem using LLMs so that customers are able to get a objective summarized view in a easily consumable format.
• ■ Moderating user provided content in real time is tricky as it takes a lot of selection and process context. Currently it is done manually which is costly, error prone and takes a lot of time. Building inhouse capability of doing multimodal content moderation by finetuning open source LLMs - 1000x faster and 10x cost effective.
• ■ Working on multiple generative AI based explorations and prototypes using top reasoning models, STT models, cursor, firebase studio, N8N etc to solve some existing customer problems related to UGC. Hands on experience on prompt tuning, RAG, finetuning smaller models (tinyllama, qwen, llama 3B) for specific usecases and explorations.

May 2022 – Aug 2024 · 2 yrs 3 mos · Hybrid

• Got opportunity to lead and solve some of the core customer facing problems of modern ecommerce collaborating with dev, business, analytics, data science, design and operation teams. Noting the top ones here.
• ■ Problem statement 1: How do you compute ideal price of products available on Flipkart solely based on customer and seller behavior on the platform and use it to influence Search ranking and relevance. Impact: Topline ₹XXXX Cr, bottomline ₹XX Cr.
• ■ Problem statement 2: In a world of inflated MRP, how to you identify actual good deals and help customers understand that with minimal cognitive load and use Search, Merchandise and Recommendation systems to surface these deals better. Impact: Topline ₹XXXX Cr.
• ■ Problem statement 3: With back to back sales, customers are mentally trained to wait for sale for purchase, even for need based products. How do you shape customer behavior to convert in BAU through detection and communication of stable pricing, alongside incentivizing sellers for great prices irrespective of sale or BAU. Impact: conversion improvement by 0.X%.
• ■ Problem statement 4: A lot of our customers don't understand technical jargons very well. How to simplify things for them in a scaled manner using generative AI. Impact: 0.X% GMV growth due to better conversion.
• ■ Problem statement 5: With too many offers on a single product, how to save businesses/brands from double discounting for a single conversion. Impact: Bottomline ₹XXX Cr.
• ■ Problem statement 6: Building basket experience for softlines (lifestyle, home, general merchandises etc) to nudge customers to buy more through incentivisation. The goal is to improve contribution margin of FK through improved scope of multiple units being shipped in a single shipment.