2025-09-18
   

Meet 'Mentor Mitra AI': A friend, a companion, a mentor every child deserves!

Have you ever wondered what is the negative impact of modern technology on children? What kind of interaction kids are having with internet and digital gadgets? Do you also feel it’s something which needs to be addressed? 

In order to solve this problem two young minds (Ankur Majumdar and Maitreya Agarwal) joined hands and innovated

‘Mentor Mitra AI’, - a friend, a companion, a mentor every child deserves.

Here is the story of young innovators Ankur & Maitreya, who won INR 3 lakhs with their project Mentor Mitra AI, an innovative educational toy designed to empower children by combining the excitement of robotics with the supportive capabilities of generative AI, proving

India's Got Intelligence

Hello everyone! My name is Ankur, and I’m a 20-year-old and currently in my third year of engineering, specialising in VLSI (Very Large Scale Integration) (a field focused on chip design and fabrication, much like what’s driving India’s semiconductor push).

Looking back, my journey into engineering was anything but straightforward. I wasn't always a fan of math in 10th grade; in fact, I initially took Physics, Chemistry, and Biology (PCB) just to avoid it.

However, my father convinced me that the medical field was too time-consuming, so I switched to PCM (Physics, Chemistry, Maths) for 11th and 12th. Even then, math was my weakest subject, and honestly, I didn't have a strong interest in engineering itself.

Like many in India, eveyone see engineering a secure path to a job rather than a passion.

My life took a turn after joining college in Noida, which was conveniently just a 20-minute drive from my home in Ghaziabad. In my first year, I started exploring the college's electronics clubs.

These clubs, run by older students and funded by the college, provided a hands-on experience that textbooks never could.

We learned about basic circuits, microcontrollers like ESP32 and Arduino, and home automation. After rigorous 9-to-5 college hours, these evening sessions were where I truly came alive.

I remember one event where we learned to simulate circuits on Proteus software and later participated in a PCB (Printed Circuit Board) design competition.

My design for an automatic street light won third prize and about ₹2,000. That moment was a game-changer. It was the first time I felt genuine excitement, realising I can turn theoretical knowledge into real-world applications.

It was clear: Project-based learning was my calling, not just bookish knowledge.

This newfound enthusiasm quickly led me to my first real-world project. Meanwhile, I also started interacting with Maitreya Agrawal, who was a topper in our batch. We didn't really have that connection in first year.

However, a series of hackathons, including the Smart India Hackathon (SIH) and our college's internal "Ride Hackathon," brought us closer.

I saw Maitreya's incredible capabilities when he almost won the Ride Hackathon with an eye-tracking Bluetooth mouse for people with physical disabilities.

He was brilliant, but frustration and working alone led him to quit before the final round.

This showed me his potential and made me think that if we teamed up, we could achieve great things, especially because there wasn't much enthusiasm for hardware projects in our institute.

But then came the genuine problem, and this time it was about children.

How many of us actually realise the impact of modern technology on our children?

Maitreya (Left) and Ankur (Right)

This was the problem we chose to solve.

We noticed that children today have extremely short attention spans, often glued to 15-to-60-second reels and YouTube shorts, making it hard for them to focus on longer content. This digital overload also contributes to issues like early-stage depression, anxiety, and difficulty forming friendships, often leading them to keep to themselves.

Furthermore, children face unique communication barriers. They might have questions they can’t ask their parents or friends due to age-appropriateness or simply feeling shy. AI tools, like ChatGPT, give generic answers regardless of a child’s age, which can be problematic, especially for sensitive topics like

"How are babies born?". A 6-year-old needs a very different explanation than a 16-year-old.

Beyond that, we recognised a huge privacy concern with popular AI tools. In tools like ChatGPT, data isn't processed on your device; it goes to their servers, often located in other countries. This poses a massive security risk, especially for sensitive personal information that users often share.

Imagine if a hacker accessed that data; it could lead to blackmailing or even compromise national security. We also noted that current AI solutions aren't always energy-efficient, consuming significant power and relying heavily on network connectivity.

Now you might be thinking, who will benefit if we solve this problem?

Our work, Mentor Mitra AI, aims to benefit children aged 6 to 16 years old. It’s designed to be a companion that addresses their unique emotional and intellectual needs.

Additionally, it will indirectly support parents, especially working ones, by providing insights into their children's activities and emotional states, allowing them to feel more connected and informed about their children's well-being.

What motivated us to solve this problem?

Our journey with Mentor Mitra AI truly began when our batch coordinator, Dr. Shruti Kalra, informed us about the e-toycathon, an initiative by MeitY (Ministry of Electronics and Information Technology) and C-DAC (Centre for Development of Advanced Computing). The hackathon’s goal was to promote in-house manufacturing and engineering of electronic toys in India, reducing our dependency on China.

By this point, Maitreya had faced several rejections in hackathons and was feeling discouraged, even sacrificing exams for a competition that didn't pan out. But something in me pushed us to try one more time. We decided to tackle the most challenging problem statement: "Mentor Mitra AI," which required building a physical AI chatbot for children to interact with.

Our mindset was, "If we're going to get rejected, we might as well go for the toughest one!".

Initially, we didn't expect much, just like our previous experiences where evaluators often lacked technical understanding. But after a round of unexpected, sharp criticism from the judges, who genuinely understood AI and hardware, we realised something crucial:  if you receive criticism, it means people are interested in your solution and want to contribute to your thinking.

This motivated us to push harder, conducting extensive research and refining our approach. The failures and challenges we faced had, in fact, prepared us for this moment, teaching us invaluable lessons in perseverance and critical thinking. We knew we had to build something that truly leveraged our electronics and AI skills, focusing on practical, on-device implementation, unlike the superficial software-only approaches prevalent in most hackathons.

Let me share – How we solved the problem?

Our solution, Mentor Mitra AI is a physically interactive AI companion designed to address the unique needs of children while ensuring privacy and efficiency.

Our prototype, developed for e-toycathon, was a 3D-printed robot with pre-fed answers for local processing, meaning no internet was required for basic interactions. For more complex questions, it could connect to Google's Gemini API with the press of a button.

We integrated context caching in Gemini, allowing us to provide "reference points" about the child, like their age, interests, and even filtering inappropriate questions, to ensure age-appropriate and personalised responses.

To make the robot engaging, we gave it motorised movements, expressive animated eyes, and four touch points on its body that triggered different emotions (happy, sad, angry). It even had a temperature sensor; if it felt cold, it would shiver, encouraging the child to wear a sweater – a subtle way to instil good habits.

The core idea was to create a friend, a mentor, who could empathise with a child and grow with them.

For the DOT 5G Hackathon, we significantly enhanced Mentor Mitra AI, focusing on on-device processing to eliminate internet dependency and further secure data.

This new version incorporates a camera that can track a child's face within a room, providing live feed to parents who might be away, and even sounding an alarm if an unknown person enters the child's vicinity.

The eyes are now more humanistic and realistic, reacting to light and showing nuanced emotions. We're also working on emotion tracking through the camera, allowing the robot to recognise if a child is sad or angry and respond with comforting songs or suggestions.

Our solution heavily relies on Edge Computing and Fog Computing principles. This means all AI computations, including LLM (Large Language Model) processing, face recognition, and emotion tracking, happen directly on the device or a local, private server.

This eliminates the privacy risks associated with sending sensitive data to foreign cloud servers, ensuring that children's information remains secure within our country and with trusted government entities like DOT.

 

 

 

Let’s get deep into the technical Details - How AI is Actually Utilised?

 

At the heart of Mentor Mitra AI is a sophisticated blend of electronics and AI. Initially, we used an AI chip with pre-fed answers for basic, offline interactions. For more dynamic responses, we integrated Google's Gemini API, employing context caching to tailor answers based on the child's profile (age, preferences, etc.) and adjusting the "temperature" parameterfor varying levels of creativity in responses.

In our advanced version, we leveraged Edge Computing and Fog Computing to bring AI processing onto the device itself. This was crucial for privacy and efficiency. We achieved this by hosting LLM models locally on private servers and fine-tuning them to our specific requirements. This approach minimises reliance on external cloud services, ensuring data stays within our secure, private network.

The robot features a Sony IMX 500 camera, which is capable of recognising up to 80% of objects in its view. We further enhanced this with our custom-trained ML (Machine Learning) model using OpenCV (Open Source Computer Vision Library) to perform face recognition for the child and their parents.

If an unrecognised person is detected, the robot can trigger an alarm. For emotion tracking, we're integrating models from platforms like Kaggle, allowing the robot to understand a child’s emotional state (e.g., sadness or anger) and respond empathetically. The speech input from the child is captured via a microphone, processed on our private server, and a tailored response is generated, all within our secure network.

For the hardware, we designed our own PCBs (Printed Circuit Boards) from scratch, rather than readily available development boards. This involved complex power management systems, boosting 3.7V battery power to 5V for microcontrollers like ESP32 and Arduino, and careful consideration of common grounding principles for motors.

We even had to learn about materials like FR4 Substrate for PCB strength and insulation, realising too late that placing a WiFi antenna directly on it would block signals. Our focus remains on creating a power-efficient system that can run ML models on minimal resources, akin to the principles of TinyML, allowing the device to operate for extended periods without constant charging.

Challenges & Failures

Early on, I struggled with academics; my JEE scores weren't great, and I got a direct admission to a branch I hadn't initially preferred. Then, in college, the biggest hurdle was the lack of enthusiasm for hardware projects. Everyone wanted to do software, app development, or website design, but electronics meant investing capital, time, and dealing with constant errors – even a simple LED blink could take hours. Seniors often had theoretical knowledge but lacked practical experience.

Maitreya, despite his talent, almost gave up after facing repeated failures and feeling that hackathon evaluators sometimes lacked the technical depth to appreciate his innovative hardware solutions.

We learned a crucial lesson about evaluators in government hackathons sometimes being unenthusiastic and just doing "core sarkari kaam" (core government work), not appreciating advanced tech.

Our own overconfidence during the e-toycathon first round led to negative feedback from judges who actually knew their stuff. This was a humbling, but ultimately valuable, experience.

The technical build for e-toycathon was fraught with errors. I designed the PCB assuming simulation was enough, forgetting basic real-world necessities like power supplies and battery management systems. I made blunders like placing the Wi-Fi chip directly on the FR4 substrate, effectively killing the signal, and forgetting the critical concept of "common grounding" for motors. These mistakes were discovered just two weeks before the final event.

During the hackathon itself, C-DAC, organising for the first time, failed to provide electronic components on time. They were relying on a third-party vendor, and the components were heavily delayed. While other teams panicked, we were lucky to have our own components we’d purchased for testing. This unforeseen delay turned into a blessing in disguise.

The late nights, fuelled by sheer determination, saw us and only a few other teams working through the night while others gave up and left.

These constant hurdles, from technical glitches to external scepticism, forced us to continuously learn, adapt, and refine our vision.

AI Readiness Skills

Our journey with Mentor Mitra AI has been a testament to the importance of what I now understand as AI Readiness Skills

  • Collaboration skills

Working with Maitreya was crucial. Maitrey’s perseverance, especially after previous failures, was incredibly inspiring. We learned to divide tasks, with Maitreya focusing on 3D printing and hardware assembly, and me on PCB design. We also sought guidance from our batch coordinator Dr. Shruti Kalra, Director Dr. Shweta Srivastava, and especially Mr. Vishnu Ram, an independent researcher and DOT official who provided invaluable, deep technical guidance.
  • Critical Thinking & Creativity:

This project demanded constant critical thinking. When our initial PCB designs failed, we had to quickly diagnose issues like power supply, common grounding, and WiFi antenna placement, finding creative workarounds like building on a breadboard. Our creativity shone in developing expressive eyes, touch-sensitive body parts, and a temperature sensor, aiming to make the robot truly empathetic and engaging for children.
  • Online Learning skills:

We relied heavily on online resources. We learned PCB design and practical electronics from YouTube channels like TechieSMS, run by Sachin Soni from Ahmedabad, an innovator we deeply admire. For advanced concepts like Edge Computing, I completed Shawn Hymel’s "Edge Computing for Microcontrollers" course on Coursera and studied the TinyML book, which teaches how to run powerful ML models on tiny, power-efficient devices. This "beyond books" approach was vital for our innovation.
  • Online Reputation skills:

We never anticipated the power of online presence. After winning multiple hackathons, I casually posted about our work on LinkedIn. I tagged 11 people, thinking I'd get maybe 100-200 likes. But within three days, it crossed 700 likes, and in just two more days, it reached over 4,000 reactions and now it’s over 7000!. This viral post showed us how quickly our work could gain recognition and inspire others, especially given the visually exciting elements like the robot's expressive eyes.

Impact So Far & Vision for the Future

Our journey with Mentor Mitra AI has yielded incredible results and instilled in us a new sense of confidence. We were surprised to win second prize and ₹1 lakh at the e-toycathon Hackathon, competing against top IITs and NITs. This win was a huge validation, showing us that our practical, hardware-focused approach could beat theoretical knowledge.

Following this, we entered the Electro Second Electronic Design Contest, an online global competition where makers upload their hardware project reports. To our astonishment, we secured second place globally, winning $1200 (close to ₹1 lakh), with Sachin Soni (TechieSMS) coming in third. Our detailed project repository, which we made open source for educational research, contributed to this win.

Most recently, we've been selected for the highly competitive DOT 5G Hackathon, which is perhaps the biggest hackathon in India. Out of 1200 initial registrations, we are among the top 50 teams who have received a ₹1 lakh seed fund. This opportunity comes with access to DOT's 5G labs and, crucially, the chance to secure IP protection for our project, allowing us to register it as a startup and file patents.

These successes, all within six months, have profoundly impacted us. We've gained immense credibility, and our college now sees us differently. We've proven that we can compete with, and even surpass, students from top-tier institutions, and that electronics, despite its challenges, offers immense potential in India, which is currently underserved in this field.

Our vision for the future is clear. With the DOT 5G seed fund, we now have the budget to further enhance Mentor Mitra AI, making its on-device AI capabilities even more robust and cost-effective. We aim to continue exploring Edge Computing and Fog Computing to ensure our solution is not only intelligent but also private, secure, and energy-efficient, directly addressing the critical concerns surrounding current AI tools. We’ve also received an invitation to present our work at the Indian Mobile Congress, Asia's biggest tech fest, at Yashobhoomi, Dwarka, which is a fantastic platform for exposure. Our ultimate goal is to commercialise Mentor Mitra AI, taking it from a successful hackathon project to an indigenous Indian product that genuinely helps children and parents. We believe that by focusing on deep tech and practical applications, we can make a lasting impact.

Your Advice to Fellow Innovators

My biggest advice to aspiring AI innovators in India, especially those like me who are in their early 20s, is simple: just start taking action. Don't wait for everything to be perfect or for formal education to catch up. Go online, participate in as many hackathons as you can – be it government-backed ones like e-Yantra or DOT 5G, or global virtual competitions like Electro. Focus on doing the work; learning will follow.

Mentor, Support & Inspiration

My journey would not have been possible without the incredible support and inspiration I’ve received.

  • Mentors: Dr. Shruti Kalra, our batch coordinator, and Dr. Shweta Srivastava, our Director, have been incredibly supportive, pushing us to participate in hackathons and advocating for our work at the institutional level. Mr. Vishnu OV Ram, an independent researcher and DOT official, provided invaluable technical guidance, sharing his deep knowledge of physical AI and 5G. His practical insights helped us refine our project significantly.
  • Support: Our college initially provided ₹8,000 for pre-testing, and the DOT 5G hackathon has given us a significant ₹1 lakh seed fund. This financial backing has been crucial for building our prototypes.
  • Inspiration: Innovators like Sachin Soni from TechiSMS have been a constant source of inspiration through their YouTube videos, teaching us practical electronics and even alerting us to global competitions like Electrow. And, on a more aspirational level, I've always been inspired by fictional characters like Iron Man, whose blend of technological genius and unwavering determination resonates deeply with me.
Got a Story?

Tell us about a story worth publishing :)
Want to Make Your School AI Ready?

Start AI Readiness Skills Mission today! We are here to help.
Want to Make Your College AI Ready?

Start AI Readiness Skills Mission today! We are here to help.
Explore AI Readiness Resources

We have created detailed toolkits, books, ebooks, videos and other resources to help schools, colleges, teachers & students become AI Ready.
Suggest a Story! If you or someone you know should be on IGI Stories, let us know.