Missing mother's breakfasts, grandmother's secret recipes- living away from home is arduous with unsatisfied taste buds. So, for an effortless life inside the kitchen, Euphotic labs, a Bangalore-based startup, came up with a solution- a culinarian robot. 

With NOSH, cooking is child's play. The AI-based assistant, which rustles up in four easy steps, can prepare about 200 one-pot dishes. 

 “We are working on automating whatever happens in the kitchen", said Yatin Varachhia, the Co-founder of NOSH, in a conversation with INDIAai. "The food you want is cooked for you and presented to you", he added. 

When the chopped ingredients are loaded, the model follows pre-installed recipes and automatically fixes a meal. A camera monitors the overall task. AI embedded in the model aids NOSH in adding ingredients in a sequence that a human would. In addition, the user can customize the taste, such as making the dish less spicy, adding less salt etc.  

Building NOSH 

The tale of NOSH dates back to 2017 when Yatin's wife started working after her MBA. Without their authentic cuisine, the Gujarati couple could not build their life in Bangalore.  

They tried and failed several means to tackle the challenges of managing cooking time with their gruesome corporate lives. Finally, Yatin researched among his friends living away from homes, and concluded that this was a fundamental problem in most households. It inspired Yatin to think out of the box and devise a generic solution.  

The inception of Euphotic Labs started when Yatin decided to automate the cooking process. He initially automated basic tasks like ingredient and spice dispensing. Then, they worked on developing elements like taste, consistency, aroma and texture.  

It took the team three years, six recipe iterations, and numerous user feedbacks to develop the final model. The first dish prepared by NOSH was a potato fry in 2018.  

AI in cooking 

A set of classifiers and object detection models were developed to analyze the cooking time for various ingredients. The team formulated the complete dataset for the model. The AI embedded in the model ensures that it functions like a human. For instance, if the recipe states that 'adding other veggies only after onions turn golden-brown', NOSH follows the step accordingly.  

When asked about the influence of AI on cooking, the team stated: 

"AI already controls what we cook in many ways. There are now models which suggest recipes just by looking at the photograph of your refrigerator". 

Being a startup in India 

NOSH wasn't built in a day. Besides being an AI-based startup, it is also a hardware company. They faced issues concerning manufacturers, suppliers, talent and many more. The rudimentary idea of the team was to build a model that could cook food that tasted good. They received a grant from the Government of India, which aided them in the early days of establishment. 

The dataset unavailability was the major hurdle the team had to face. They visited a village and asked the local homemakers to cook while they collected data. "Data collection continues to be one of the biggest challenges that we face today", stated Yatin. Also, executing the process in real-time was difficult for the team as their computer's processing power was low.  

Four years after its invention, NOSH is in a good place. "We have a good number of pre-orders for our product", added Yatin. 

According to the NOSH Co-founder, two kinds of chefs are those who create new recipes and others who follow existing recipes. He believes that, in the long run, AI will be able to invent novel recipes with the available taste palette.