Privacy first robotic companion running edge AI locally. Combines vision, voice & emotion detection, no cloud dependency, 101% privacy.

From an Idea to an Intelligent Robotic Mentor

The journey of Mentor Mitra AI began with a simple yet powerful question:

Can we create an AI-powered mentor that helps children learn, encourages curiosity, understands emotions, and most importantly, protects their privacy?

Today's AI assistants are incredibly powerful, but most of them rely on cloud servers. Every conversation, question, and interaction is sent over the internet. For children, this raises important concerns around privacy, accessibility, and dependence on continuous connectivity.

To address these challenges, we set out to build Mentor Mitra AI, a privacy-first Edge AI robotic mentor that can see, listen, understand, and respond locally without relying entirely on cloud infrastructure.

Components Used -

Raspberry Pi 5 8GB

Single Board Computer 2.4GHz 4 Core 8GB RAM X 1

Raspberry Pi Camera V2

Raspberry Pi Camera V2 X 1

LDR -Photocell Photoresistor

LDR -Photocell, Photoresistor X 1

Servo Motor MG995

Servo Motor MG995 X 4

Sound Sensor Module

Multiple Function Sensor Development Tools X 1

5 W, 8 Ohm, Speaker

Speakers & Transducers 102 X 37 mm, Square X 1

Microphone

Microphones microphone, 4 mmx1.5mm X 1

MPR121 Proximity Capacitive Touch Sensor Controller

MPR121 is a proximity capacitive touch sensor X 2

Lithium Ion Battery 3.7V 2500mAh 18650

Consumer Battery Photo Battery 3.7V 2000maH X 2

LM2596 Adjustable Buck Converter

Power Management IC Development Tools LM2596ADJ X 2

Connecting Wire Jumper Wires

Connecting Wire Breadboard wires

Esp 32 board

It's and Microprocessor board X 2

round TFT

GC9A01 Driver X 2

Before building the hardware, we carefully planned the overall architecture of the system.

We wanted Mentor Mitra to combine multiple technologies into a single platform:

1. Conversational AI
2. Computer Vision
3. Emotion Recognition
4. Robotics
5. Voice Interaction
6. Edge Computing

The architecture was designed to be modular, allowing individual components to be upgraded independently while keeping the overall system scalable.

The Raspberry Pi 5 was chosen as the primary computing platform due to its balance of performance, flexibility, and community support.

The Raspberry Pi acts as the brain of Mentor Mitra AI and is responsible for:

1. Running AI services
2. Processing camera feeds
3. Managing voice interactions
4. Coordinating robotic movements
5. Handling communication between different subsystems

By using a powerful edge-computing platform, we ensured that most intelligence remains close to the user.

Since we wanted a unique and customizable design, we created the complete enclosure using Autodesk Fusion 360.

Fusion 360 allowed us to rapidly prototype ideas, design mounting points for electronics, create cable routing channels, and ensure all components fit together correctly before printing.

The robot was divided into multiple printable parts:

1. Base.stl
2. Body_Main.stl
3. Body_Back.stl
4. Front_Face.stl
5. Head_Main.stl
6. Left_Arm.stl
7. Right_Arm.stl

Several iterations were developed before arriving at the final design. Particular attention was given to:

1. Child-friendly appearance
2. Structural strength
3. Ease of assembly
4. Internal electronics mounting
5. Future expandability

This CAD-first approach significantly reduced fabrication time and allowed mechanical issues to be solved before manufacturing.

After completing the CAD design, all enclosure components were manufactured using a 3D printer.

The modular design made printing easier and allowed individual parts to be reprinted whenever modifications were required.

The printed components include:

1. Main body enclosure
2. Robot head
3. Arms
4. Face panel
5. Base structure
6. Servo mounting components

The flexibility of 3D printing enabled rapid prototyping and helped transform a collection of electronic modules into a complete robotic product.

Most importantly, it enabled us to create a unique identity for Mentor Mitra AI rather than building another generic robotics platform.

Design Objectives

1. Child-friendly appearance
2. Modular construction
3. Easy maintenance
4. Lightweight structure
5. Future expandability
6. Rapid prototyping capability

The final design successfully integrates electronics, sensors, displays, and mechanical systems into a cohesive robotic platform while maintaining a compact and visually appealing form factor.

Since Mentor Mitra was designed to be portable and operate independently of wall power, an efficient battery-powered architecture was a key design requirement from the very beginning.

The system uses two separate power sources to ensure stable operation of both the AI computing platform and the robotic movement system.

Raspberry Pi Power System

The Raspberry Pi 5 serves as the primary AI processing unit and handles computer vision, speech recognition, conversational AI, and system coordination. Due to its higher power requirements, the Raspberry Pi 5 is powered using a 45W USB-C power bank, allowing the robot to operate completely untethered while providing sufficient power for AI workloads.

Servo and ESP32 Power System

The robotic movement system is powered by a dedicated rechargeable lithium-ion battery pack built using two 18650 cells, each rated at 3.7V and 2000mAh.

The cells are connected in series, producing a nominal output voltage of 7.4V. Since the MG995 servo motors operate most efficiently around 6V, a DC-DC step-down converter is used to regulate the battery voltage from 7.4V down to a stable 6V supply.

The same battery system also powers the ESP32 microcontroller responsible for servo control and peripheral management.

Mentor Mitra uses a total of four MG995 servo motors:

1. Two MG995 servos form a pan-tilt turret mechanism that controls the robot's head movement for face tracking and natural interactions.
2. Two additional MG995 servos are used to control the movement of the robot's arms, enabling expressive gestures and animations.

To simplify charging and improve usability, a dedicated lithium-ion charging module has been integrated into the battery system, allowing the rechargeable 18650 cells to be safely recharged without removing them from the robot.

Why Separate Power Sources?

Separating the AI processing power supply from the motor power supply provides several advantages:

1. Improved system stability
2. Reduced voltage fluctuations caused by servo movement
3. Reliable Raspberry Pi operation
4. Better servo performance and torque delivery
5. Easier power management
6. Extended battery life

This dual-power architecture allows Mentor Mitra to function as a truly portable robotic companion, capable of operating in classrooms, exhibitions, hackathons, workshops, and learning environments without requiring a direct connection to wall power.

A mentor must be able to communicate naturally.

To achieve this, we developed a complete voice processing pipeline.

When a child speaks, the audio is captured through a microphone and converted into text using speech recognition technology. The text is then processed by a locally hosted AI model, which generates an appropriate response. Finally, the response is converted back into speech and played through the speaker.

This creates a natural conversational experience that feels similar to interacting with a real mentor.

Voice Interaction Flow

Child Speaks → Speech Recognition → AI Processing → Response Generation → Speech Output

Human interaction is not limited to speech. Facial expressions, eye contact, and gestures all play an important role in communication.

To make Mentor Mitra more interactive, we integrated an AI camera system capable of:

1. Detecting faces
2. Tracking users
3. Recognizing emotions
4. Understanding visual surroundings

The camera continuously captures frames which are processed by computer vision models running on the edge.

This allows the robot to understand who it is interacting with and respond more naturally.

One of the key features of Mentor Mitra is its ability to maintain eye contact with the user.

Using real-time face detection, the robot identifies the position of a person's face and continuously adjusts its orientation using servo motors.

This makes interactions feel significantly more engaging compared to static devices.

Servo Turret - Controlled through Battery Powered .

Check The code for implementing on Raspberry Pi 5

To make Mentor Mitra feel more alive and approachable, we designed an expressive eye system using circular TFT displays.

The eyes are capable of:

1. Blinking
2. Looking around
3. Showing attention
4. Simulating emotions
5. Following user movement

These small visual cues dramatically improve engagement, especially when interacting with children.

Instead of appearing like a machine, Mentor Mitra begins to feel like a companion.

These are Light Responsive by connecting to an LDR (Light Dependant Resistor) -

Understanding emotions is an important part of effective mentorship.

We implemented emotion recognition capabilities that analyze facial expressions and identify emotional states such as:

1. Happy
2. Sad
3. Neutral
4. Surprised
5. Confused

The detected emotion influences how Mentor Mitra responds.

For example:

1. A confused child may receive additional explanations.
2. A frustrated child may receive encouragement.
3. A happy child may be rewarded with positive reinforcement.

This enables more personalized and empathetic interactions.

https://github.com/ageitgey/face_recognition

The intelligence layer serves as the decision-making core of the system.

We integrated conversational AI capable of:

1. Answering questions
2. Explaining concepts
3. Holding conversations
4. Encouraging curiosity
5. Adapting responses for children

Unlike conventional assistants that rely entirely on remote servers, our long-term vision focuses on edge-first AI deployment to reduce latency and enhance privacy.

The AI is designed to act not only as an assistant but also as a learning companion.

Basically Used LLama3.2:3b Open source LLM Model -

Building Mentor Mitra AI presented several technical challenges.

1. Real-Time Processing
2. Running vision, voice, AI, and robotics simultaneously requires efficient resource management.
3. Privacy Preservation
4. Maintaining intelligence at the edge while minimizing cloud dependency required careful architectural decisions.
5. Human-Centered Interaction
6. Creating a robot that feels approachable and engaging is much more difficult than simply building a functional system.
7. System Integration
8. Integrating multiple technologies into a single platform demanded extensive testing and debugging.
9. Each challenge provided valuable learning experiences and pushed us to innovate further.
10. Results and Impact

Mentor Mitra AI successfully demonstrates:

✅ Conversational AI interaction

✅ Face detection and tracking

✅ Emotion-aware responses

✅ Expressive robotic behavior

✅ Edge AI capabilities

✅ Privacy-focused architecture

✅ Educational assistance for children

More importantly, it demonstrates how intelligent educational companions can be built while prioritizing privacy, accessibility, and meaningful human interaction.

Looking Ahead -

Mentor Mitra AI is not just a prototype—it is a step toward a future where AI serves as a trusted educational companion.

Future developments will focus on:

1. Multi-language support
2. Personalized learning journeys
3. Enhanced emotional intelligence
4. Improved robotics
5. Classroom deployment
6. Rural education accessibility
7. Advanced Edge AI capabilities

Mentor Mitra AI represents our vision of what educational technology can become when robotics, artificial intelligence, and human-centered design come together.

By combining conversational AI, computer vision, emotion recognition, and robotics into a single platform, we have created a system that not only teaches but also engages, understands, and inspires.

We believe the future of AI should empower children, protect privacy, and make learning more accessible for everyone. Mentor Mitra AI is our contribution toward that future. 🚀🏆