Blinkit understands that the question “Aaj kya khana hai?” (What’s for dinner today?) can be a daily challenge for many households. To address this, Blinkit is excited to introduce ‘Recipe Rover,’ an AI-powered solution that helps you overcome this dilemma. With Blinkit’s AI Recipe Generator, users can explore new flavors, expand their culinary horizons, and enjoy delicious meals without the hassle of deciding what to cook each day.
Recipe Rover, a collaborative effort by a team of engineers and a project manager at Blinkit is a powerful engine that combines ChatGPT and Midjourney technologies. This innovative solution is designed to suggest recipes and assist users in finding the necessary ingredients for their culinary adventures. Blinkit’s AI Recipe Generator aims to provide a seamless and user-friendly experience for discovering recipes and simplifying the process of gathering ingredients.
In a blog post by Vaibhav Bhutani, (which was originally posted by Albinder Dhindsa, CEO of Blinkit) the AI product lead at Blinkit, he highlights the brand’s goal of making instant recipes as prominent and convenient as instant delivery. Blinkit’s AI Recipe Generator ensures that only recipes with readily available ingredients are shown to users. However, developing this level of convenience was a time-consuming process, as it took them approximately 3-4 hours to build each recipe page. Despite the challenges, Blinkit is dedicated to providing users with a seamless experience when accessing and preparing recipes.
Idea Behind Blinkit’s AI Recipe Generator
Blinkit wanted to make instant recipes as good a feature as instant delivery. Not only that, Blinkit aimed to only show customers recipes they could make right away, considering they had all the necessary ingredients available. However, building this level of convenience proved to be time-consuming, requiring 3-4 hours to create just one recipe page.
Blinkit also faced limitations in understanding cuisine and event-specific recipes, particularly regional and festival-specific ones. This resulted in missed opportunities due to long lead times and lacking context.
To address these challenges, Blinkit introduced the AI Recipe Rover! With Recipe Rover, Blinkit now has an extensive repository of recipes accessible through the search bar, along with a curated list of products needed to cook them. It’s important to note that the corresponding recipe won’t be shown if Blinkit can’t assist with the ingredients.
Also Read: Zomato-Blinkit and Never-ending Trend
So, how does this benefit Blinkit’s customers?
- Simplified meal planning: Customers have access to a range of simple recipes with ingredients that can be quickly delivered, all in just a few clicks.
- Occasion-based curation: Whether it’s mango season or Ganesh Utsav, Recipe Rover provides curated recipes to suit specific events and festivals. For example, customers can head to the mango recipe corner during mango season or make modak at home for Ganesh Utsav.
- Nutritional awareness: Each recipe includes information about the macro-nutrients contained in the meal, helping customers make healthy choices.
This is where Blinkit’s AI-enabled Recipe Rover comes into play. It generates thousands of recipes within a couple of hours, enabling Blinkit to cover a wide range of recipes tailored to ingredients, events, and festivals at the push of a button.
How did Blinkit build the entire engine for Recipe Rover?
The Game-Changing Impact of GPT on Recipe Development
The team at Blinkit experimented with Chat GPT to curate recipes and was astounded by the quality of AI-generated recipes. Leveraging the power of OpenAI’s GPT-3.5 and GPT-4 models, they generated an impressive 50,000 recipes in one go, significantly expanding their recipe bank.
To better understand customer preferences, the Blinkit team delved into their own data and refined their knowledge accordingly. This allowed them to cater to each customer’s unique tastes and create a dynamic culinary experience that continuously evolves. Additionally, the Search team at Blinkit worked diligently to optimize the use of these APIs, refining prompts that facilitated the construction of JSONs that seamlessly integrated into their existing elastic search system.
Crafting delicious recipes: Using Midjourney
To achieve this, the people at Blinkit undertook the task of fine-tuning their own Stable Diffusion model using a vast dataset of over a million images. Through meticulous training, they taught the AI model to recognize and represent the unique characteristics of different dishes. This included imparting knowledge on the visual aspects of desi chowmein, distinguishing between the textures of paneer and tofu, and more.
To truly captivate their customers, the team at Blinkit understood that they needed more than just well-written recipes. With the help of Stable Diffusion 1.5 and Midjourney techniques, they successfully crafted visually appealing images for each recipe. These stunning visuals perfectly complemented the AI-generated text, offering a tantalizing glimpse of the final dish and capturing the attention of their customers.
Creating attractive and achievable recipes: Utilizing vector search
Vector search is a vital component of Recipe Rover, enabling the team at Blinkit to offer customers an efficient method of discovering the necessary ingredients for their recipes.
By extracting ingredients from the AI-generated recipes and aligning them with the available products on Blinkit, vector search ensures that customers are presented with recipes they can prepare with the ingredients they have on hand. This intelligent matching of ingredients simplifies the process for customers, allowing them to find all the necessary ingredients in a single search. Through the implementation of vector search, the people at Blinkit have enhanced the user experience by providing a seamless and convenient way for customers to access recipes tailored to their available ingredients.
Presenting carefully curated recipes: Dynamic search and personalized recommendations
When customers search for popular keywords, Recipe Rover, developed by the team at Blinkit, displays recipe image cards as part of the search results. These visually appealing cards present comprehensive information including ingredients that can be ordered, nutrition details, and cooking directions.
Also Read: Swiggy Dineout Launches Its First Ad Featuring Farhan Akhtar and Shibani Dandekar
To further enhance the user experience, Recipe Rover incorporates an auto-suggest feature that helps customers discover new dishes based on their preferences. This intuitive functionality suggests recipes that align with their culinary interests, allowing for exploration and experimentation.
Blinkit has taken the eternal question of “Aaj kya khana hai? Aaj kya banana hai”? to a whole new level of seriousness and quirkiness. With the help of AI, they have conjured up a colossal panel of recipes that will send foodies into a frenzy of gastronomic delight. Brace yourself for a culinary adventure like no other as Blinkit unveils an extraordinary collection of delectable recipes that will tantalize your taste buds and leave you craving for more. Get ready to embark on a delightful journey of flavors and indulge in the ultimate foodie extravaganza
