Radhika Kukreja · Product Manager

I picked up popular apps, found the most relevant problems they have, and tried to solve them.

A small collection of AI projects built on surfaces people use every day — Flipkart, Swiggy Instamart, and a few quiet internal tools.

• 01Swiggy Instamart × Instagram

  Instagram Reels to Instamart Cart

  Problem

  200M+ people watch recipe reels every week. They save them, then forget by the time they open a grocery app. Intent and purchase never meet.

  What I built

  A Chrome extension (and a native share-target vision) that extracts ingredients from a reel with Gemini and surfaces them inside Instamart — deduped, brand-matched, pantry-aware.

  Chrome extensionGeminiNative share flowMetrics framework
  Read the case study
• 02Flipkart Compare

  Smart Compare — decide faster

  Problem

  Buyers can't decode phone spec sheets. They leave the Compare page to watch YouTube reviews, and purchase intent stalls.

  What I built

  Ask for intent first (gaming, photos, calls), then translate every spec row into a plain-language outcome — flagging where the difference is meaningful and where it isn't.

  Intent captureAI explanationsCompare-page conversion
  Read the case study
• 03Internal QA workflows

  AI agents for QA error & feedback tracking

  Problem

  QA signals — bug reports, user feedback, product info drift — get lost across Sheets, Gmail and Slack. Triage is slow and inconsistent.

  What I built

  A small fleet of agents that watch the surfaces, log structured entries to Sheets, and route the right messages to Gmail and Slack — so the team responds in minutes, not days.

  Visual Web InspectorQA Error TrackingQA Feedback TrackingAuto Update Product Info