A proper diet is an important part of leading a fit healthy life. The food we eat directly affects our health and fitness. Therefore, it is important for everyone to eat proper food and also not compromise on the taste. Especially during this Coronavirus pandemic, where people are mostly staying at home, not being able to go out much especially to places like gyms for working out, the obesity rate increases. Obesity leads to a lot of health-related problems. So, people must follow a healthy diet regularly in order to avoid obesity but not compromise on the taste of the food at the same time.

We came across a diet called the Keto diet that reduces the sugar intake and burns the body fat for the regular functioning of the body. So, people can have their favorite foods that are rich in fat and have very less sugar content, and stay healthy at the same time. But there was no proper guide for it, the articles over the internet were too complex for a common person to understand and did not include the products that the people reading them could directly buy from the market. So, the people following the Keto diet have a hard time searching food products, looking into the description of the product's contents, and comparing it with the Keto safe content limits that they read on the internet blog when they go shopping. This would take more time for the people following the Keto diet for shopping, cooking, and ordering food at a restaurant. Also, the diet consultants charge a lot for a limited time of consulting.

Therefore, we wanted to make something that people can use any time, get friendly suggestions that do not have too much nutrition jargon, talk to it just like the way they talk to their diet consultant, and use it easily for free.

What it does

KetoWit is a voice-based Keto diet assistant that helps people following the Keto diet with shopping, deciding what to cook or order at restaurants. It consists of three main functions (also called intents) for now:

  1. It can help the user to check if a specific food product, dish, or ingredient falls under the keto diet.
  2. Suggests Keto-friendly food products and brands that the user can shop for at the supermarket.
  3. If the user wants to know the nutritional value of a specific food product, ingredient, or a dish, it can tell the user about it.

In this way, people who do not know much about nutrition and do not have much time for looking into food specifications can also use this for choosing what to have.

How we built it

At first, we found a dataset with a lot of food products, dishes and ingredients, along with their nutritional value. After that, we filtered all the food products, dishes, and ingredients that were Keto diet-friendly. Then, we uploaded 1000 of those filtered products to our hosted PostgreSQL database on the Heroku cloud platform.

After that, we created and trained the three intents on and also built a Django-based web app. We integrated the WebKit speech recognition module in the front end so that whenever a user speaks something into the front end, it converts the speech to text and sends it to our Django-based back end. The Django-based back end would then make a request to with the user's speech as the utterance in order to get the intent and the food entity name. Based on the intent, the Django app would execute a specific function that would query our food database with the food entity recognized from the user's speech to get the results. Then, the results would be put into a sentence form and sent back to the front-end where the Text to Speech engine would read out the result to the user. Later, we deployed the web app to Heroku.

Challenges we ran into

  1. We had to research thoroughly into the Keto diet and the nutrition attributes associated with it.
  2. Dataset was hard to collect and process because of the huge number of attributes.
  3. The Heroku cloud platform had limits on the database and app usage.

Accomplishments that we're proud of

Finishing our first project that uses a voice-based assistant whose utterances are trained on, the number of lives that this project can impact, the ease of use that we achieved.

What we learned

  1. Creating a voice-based assistant fast and easy.
  2. Simplifying user experience.
  3. Using and facebook's developer platform.
  4. Making AI-driven technologies that have an impact.
  5. Using open-source and other free to use tools extensively in order to build impactful technologies.

What's next for KetoWit - A voice-based Keto diet assistant

We want to make it personalized for every user using it with options for a shopping to-do list, scheduling, diet tracking, and vegan specific diet recommendations. We also want to add more dishes, ingredients, and food products into our database so that the users can look up a wider range of food items.

Share this project: