Published in


Solving for Indian Context— Deeptech challenges

Photo by Hitesh Choudhary on Unsplash

The drivebuddyAI platform is built with the fusion of EdgeAI models which does the inference and CloudAI models which learns from the data and update the EdgeAI model on the device with the updated learning.

The core idea of drivebuddyAI lies in improving the behavior of the drivers and as a result encouraging them to make the driving ecosystem safer and better. The beahvior derivation comes from the driving patterns and that can be learned from by creating the context of the surrounding of the vehicle. The alerts which are meant for assisting drivers in case of any critical collision like situations are also a powerful data to understand that based on our built intelligence how human drivers are reacting to it. We are building AI systems which are meant to be used by humans to assist them in different tasks and for that the AI also has to learn about the task as well as it has to be adjusted according to the human for whom the intelligence is going to be used.

Generalizing in AI means it can adapt to the different situations/scenarios, act according to the learning and to achieve that it has be trained with all the wide variety of the data points across the world.

Our AI team at @drivebuddyAI comes across lots of challenges while building the product which can successfully be able to interact with human drivers and also for human drivers to listen to them and respond.


Consider a scenario where we have an AI model whose job is to detect different categories of objects on the road. India is the place with wide and wild variety of types of vehicles, their color schemes, their shape etc and there are lots of similar features but there are lots of differences when you change the city. This comes up with the huge challenge of identifying the right kind of object for giving the alert at the right time to assist.

Dumper truck identified as Tempo
Left. Tempo identified as Truck, Right. Truck identified as Tempo

Due to the similarity of the features in Truck and Tempo categories of the vehicle, they are often mis-classified. The solution to this is to go deep into the data and optimize the models to avoid these detections. In a general concept, somebody can say that whatever the class may be, for a device the important thing is to detect the object so that alerts can be generated. The argument might be correct as well but consider, tempo being the light vehicle still is not as dangerous as truck for a car driver. Also to identify the behavior of a driver according to the situations, it is important for a model to classify objects in a right way.


Consider a scenario where the vehicle is being driven at the night time. AI models are mostly get tested in such situations.

Not all the problems can be solved by just training the model

To optimize the model the data has to be understood really well and as an AI engineer working to solve the problem, it becomes crucial to learn the data, different scenarios and also the data source like what kind of camera is being used, how it behaves in low light conditions and conditions where the vehicle is moving out of the tunnel where the chances of image getting saturated is higher. Here’s one such scenario where drivebuddyAI team is working out the solution not only in AI but in conjunction with Hardware, Computer Vision and then AI.

Glare from the opposite direction vehicles make them invisible behind light beam
Both the images shows that how lighting in the dark environment impact the AI detection

Solving an AI problem in an uncertain and challenging environment takes a lot of patience, lots of efforts and hardware as well as a right approach. Combining all these together and working out a solution with Hardware, Computer Vision & AI can solve the challenge. All it takes is a thought process required to solve unknown challenges & problem statements.

India is considered to be one of the most difficult place for driving for humans and that’s why it is assumed that India will be the last place for getting an autonomous vehicles. I believe we can prove that wrong as well as taking up the challenge to solve the AI problem for the Indian context can solve majority of the world problems.

Join the team

Do you think our problems are different but worth solving? Do you think the problems we are working out are complex and requires a right kind of approach with depth of deep-tech & computer vision. We are looking for the talent to work on such problems with us when are building from India to solve the problems of the world by taking on the challenge.

Apply Here




Building an AI nervous system, that learns human behavior to augment their decision-making capabilities for empowering the mobility ecosystem

Recommended from Medium

Artificial Intelligence vs Robotics: All the Facts You Need to Know

How can AI-Mediated Conversations Impact our Capacity to Interact?

The AI that speaks the language of chemical reactions natively

Is technology disrupting or expanding the human workforce?

Overview | Betting on Red and White

The Top 10 AI Start-UPS to Watch

Aidrivers and Jacobs enter new agreement to advance AI-enabled autonomous solutions in industrial…

Why Companies Need To Improve English Language Testing

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store


Entrepreneur, Startup Founder & CEO @drivebuddyAI. Learning from human drivers for the future drivers.

More from Medium

Electron Refraction & Reflection for Synchronous Oscillatory Systems

A decade of modern Artificial Intelligence — from sci-fi to everyday commodity.

AI has become a commodity that touches our lives whether we are fully aware of it or not!

COVID-19 crowd analysis — Analyzing crowd density using computer vision

Face Recognition DCNN