AI Driven Interactive Learning Platform: A Systematic Approach to

Personalized Learning and Persona-Based Content Delivery

Niteen Dhutraj, Kalparatna Mahajan, Jayesh Bhandarkar, Dhanashri Chaudhari and Jagruti Wagh

Department of Information Technology, SVKM’s Institute of Technology,

Dhule, Maharashtra, India

Keywords: Adaptive Learning Platform, AI-Powered Learning Platform, GenAI, Interactive Learning, Large Language

Models, Persona-Based Education, Personalized Learning, Retrieval-Augmented Generation.

Abstract: Education is evolving with the integration of artificial intelligence, offering opportunities for enhanced

personalization and learner engagement. However, many learners struggle with finding tailored support that

matches their individual preferences and motivations. This paper presents an AI-powered interactive learning

platform that uses information and character models for creating effective and interactive learning scenarios

related to different real and fictional personalities. Constructed to maintain interest based on typical and

motivational modes of learning, all the personas regulate their answer choices based on learner type, mode of

learning and preferences all in real time. The system introduces adaptation based on a Large Language Model

trained in various data types, thus optimizing engagement, satisfaction and learning retention. In terms of

retrieval-augmented generation, it dynamically responds to users being capable of providing answers from

apparently diverse personality characteristics and of teaching methods for the different learning styles of the

majority of users. AI features of the system include adaptability, engagement profiles and response analytics

that makes the educational environment both engaging and continuously responsive. This work examines the

creation and deployment of persona-based AI in learning environments while also highlighting how the

technology could improve teaching and training experiences with intelligent and adaptive communication

skills.

1 INTRODUCTION

In recent years, AI has significantly influenced digital

learning enabling learners to realize the needs of the

learning process. In most representative conventional

online learning models, there is no or limited

interaction between the student and contents; as such

the content delivery is mostly static and may not be

friendly enough to suit individual learning needs.

This causes a problem as learners cannot relate to

most content they come across, their motivation,

retention and overall learning experience is

negatively affected.

This paper introduces an AI-driven learning

approach that utilizes a persona-based interaction

strategy are proposed to increase consumer

interactions and satisfaction in digital learning

environment. As with the application representing

well-known personalities and fictional characters:

Virat Kohli, Shah Rukh Khan, Doraemon and Steve

Jobs, the application adapts to the requirements of

students of different ages and knowledge levels. This

approach is interesting and realistic because each

persona has distinct attributes and ways of

communication. The system employs a LLM that has

been fine-tuned to use various datasets to make

responses for retrieval and generation that vary

depending on the user’s inputs, to give real-time,

relevant feedbacks in accordance with the learner’s

preference (Gan, Sun, et al. , 2019).

The objective of this research is to explore the

potential of persona-based AI in the contexts of

learner engagement and personalization. While

adaptive learning platforms have shown promise,

they often lack contextual personalization and

engagement, leaving learners disconnected. Our

persona-based approach addresses this gap by

tailoring not only content but also interaction styles,

creating a more relatable and immersive learning

experience. Implications from this study provide an

484

Dhutraj, N., Mahajan, K., Bhandarkar, J., Chaudhari, D. and Wagh, J.

AI Driven Interactive Learning Platform: A Systematic Approach to Personalized Learning and Persona-Based Content Delivery.

DOI: 10.5220/0013595400004664

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 3rd International Conference on Futuristic Technology (INCOFT 2025) - Volume 2, pages 484-492

ISBN: 978-989-758-763-4

understanding of how persona-driven AI may help

learning styles and outline the development of

subsequent, digital learning environments suggests

the possibility of change through persona-driven AI

learning.

In Section I, we introduce the AI-powered

learning platform, designed to enhance educational

engagement through persona-driven interactions.

Utilizing RAG and LLM model, the system

dynamically adapts responses based on distinct

personas, providing users with tailored learning

experiences (A. G, 2024). The rest of the paper is

organized as follows: Section II reviews related

literature on AI-driven personalized education. In

Section III, we discuss the methodology, including

the system architecture, LLM fine-tuning and

persona-specific adaptations. Section IV covers the

analysis of results and Section V concludes the

research study.

2 LITERATURE REVIEW

In research paper (Rizvi, 2023), the study focuses on

AI-enabled learning systems that investigate

alternative frameworks and models for language and

programming instruction, employing Bayesian and

neural network approaches. However, many of these

technologies are still in development and have

minimal commercial potential. One major criticism is

that it does not meet advanced learning needs, which

leaves clients struggling with the technology. The

project's purpose is to overcome these gaps by

analyzing existing AI systems and inventing novel

classroom applications, resulting in increased student

engagement and learning outcomes.

According to Baillifard et al. (Baillifard, Gabella,

et al. , 2024), AI-powered tutoring systems can

potentially improve customised learning, user

engagement and academic performance. Immediate

feedback, spaced repetition and dynamic engagement

are all effective methods for capturing student

attention. However, problems arise when AI lacks

sufficient personalisation, leading in disengagement

and poor learning outcomes. According to study,

platforms that do not deliver real-time feedback or

engaging conversational interfaces do poorly. AI-

powered celebrity lecturers, for example, provide a

way around these limits, potentially increasing

engagement and improving educational

achievements.

In(Attigeri, Agrawal, et al. , 2024), Girija Attigeri

et al. presents a chatbot for engineering college

admissions was developed to handle a high volume of

queries from students and parents throughout the

counselling process. It accurately responds to client

questions utilising pattern matching, TF-IDF

vectorisation and neural networks. This model,

known as Hercules, outperformed others because to

its sequential modelling and optimisation approaches,

which provided 24-hour support, combated

misinformation and improved user experience.

In (Eyupoglu, Korkut, et al. , 2024) researchers

evaluated the effectiveness of two AI chatbots, Bard

by ChatGPT-3.5 and by administering a 90- query

survey to 46 emergency medicine residents. While

Bard scored 55.5% with 50 correct answers,

ChatGPT-3.5 achieved 60% accuracy by answering

54 questions correctly. While both models did well in

endocrine illnesses, they struggled with digestive

problems and ECG interpretation. These results

highlight the need for additional study to improve AI

models' effectiveness in emergency medical

scenarios and cast doubt on their trustworthiness in

medical education.

McGrath C. et al. in (McGrath, Farazouli, et al. ,

2024, presents the impact of chatbots driven by

generative AI on procedures in higher education was

examined. The work adds to the theoretical body of

work to correct for tendency of prior literature to

hyperbolise the potential of these technologies. As

much as chatbots provide for the learners, some

questions arise as to how they impact equity and

learners’ engagement across learning communities.

The authors suggest using the form of qualitative

research so that the difficulties and changes

experienced by GAI chatbots and related educational

stakeholders can be better understood clearly.

The research conducted by Shahri et al. (Shahri,

Emad, et al. , 2024)focused on evaluating the

potential of GPT-4 as an AI-powered tutoring system

for delivering personalized education. Their work

aimed at exploring the possibility of applying the

concept of the advanced language model, GPT-4, to

educational practice. The researchers aimed to bridge

the gap for the effectiveness of the system in

delivering tutor-specific instruction in a student-

centered approach and in delivering feedback in real-

time. Their work was focused on trying to bring the

traditional education paradigm closer to the new age

personalized learning based on AI technology and the

main goal was to advance education through

technology (AlShaikh and Hewahi, 2021).

Frank et al. (Frank, Herth, et al. , 2024 has

discussed the effectiveness of Large Language

Models to design an Intelligent Tutoring System for

R programming education for teaching learners at

AI Driven Interactive Learning Platform: A Systematic Approach to Personalized Learning and Persona-Based Content Delivery

485

personalized level with feedback and learning path

preference. Their studies compared different

language models as a way of identifying the most

suitable strategies towards the deployment of

artificial intelligence enabled programming tutoring

system (Yesir, Rawat, et al. , 2023).

3 METHODOLOGY

The intervention of this research is to develop an AI

based interactive learning platform as illustrated

below in figure 1 which will help in making online

learning more personalized and engaging. The

platform leverages a MiniLM transformer for

vectorization and integrates LLM for content

transformation techniques. The system produces the

course modules according to the user queries and then

applies the personas to amend the modules. The

methodology is entirely based on the fusion of

technologies like Django for the backend, React for

the front end or the user interface, while LLM namely

Google’s Gemma Model as the power that steers the

AI-based conversation. This combination makes the

system highly scalable and easy for the users to

navigate while at the same time being capable of

handling highly complex learning queries (Makharia,

et al. , 2024). To have a thorough understanding of its

functionality, we will look at the system architecture,

content generation process, data embedding

strategies, LLM fine-tuning and evaluation strategy

for determining its impact on learning results. The

architecture diagram, in figure 1, illustrates the

components and flow of data within the System,

designed to create a persona-based, interactive

learning experience for users. This architecture

integrates key modules for content processing,

personality modeling and interactive user response,

aiming to deliver an AI-driven educational assistant

that is both informative and engaging (Trivedi, , et al.

, 2023). The steps in this process are outlined below.

3.1 Input Module

In the initial stage, users enter the desired topic

through a chat interface as well as audio input. This

input is processed using Natural Language Processing

algorithms, which identify the requested topic in the

system. The input is mapped to the available topics in

the database using a semantic similarity model. If the

topic exists, the system progresses to generate

relevant learning modules.

Equation 1: Semantic Similarity Score

Calculation

The similarity between the user's input and stored

topics is calculated using cosine similarity:

Similarity Score=

∑













∑











∑











(1)

Where A and B are vector representations of the

user's input and stored topics, respectively. This

ensures that the topic chosen by the user matches one

of the pre-defined subjects in the system. An interface

seamless topic identification.

Figure 1: System Architecture

INCOFT 2025 - International Conference on Futuristic Technology

486

3.2 Database Module

The project involves collecting large volumes of data

from podcasts, social media, speeches and movie

dialogues, among other activities, to create a strong

knowledge base. This data is employed to predict

personality and to fine-tuned the LLM for the given

types. This helps to the development of an interactive

learning platform that reacts based on the

characteristics laid down. The content database is

structured into three main layers: topic categories,

specific modules and AI tutors. As for the learning

modules of the selected topic, the listing is retrieved

from the database. Each AI tutor is pre-trained with

distinct response models, reflecting the persona of the

tutor. The database also maintains the record of the

conversation such as log, history etc to enhance user

related data over the period as shown in figure 2.

Every tutor is fine-tuned using Google’s Gemma

Model and the conversation that the user had with the

tutor will be recorded for future use.

Equation 2: Data Retrieval Query

The system queries the database to retrieve modules

based on user-selected topics:

Module Set = {M





,....,M



} (2)

Where M is the i

learning module associated with

the chosen topic.

Figure 2: Database Structure

3.3 AI Tutor Selection Module

Users select their AI tutor from a list of available

personas, each fine-tuned to respond with a specific

style. The system utilizes a Transformer-based

architecture for generating responses in real-time.

After selecting a tutor, the fine-tuned model

associated with the tutor is loaded and the

conversation begins.

For instance, when a user selects the 'Virat Kohli'

persona and queries about 'effective leadership,' the

system uses semantic similarity (Equation 1) to

retrieve domain-relevant content, followed by the

persona-specific response generation using fine-

tuned GPT models. The response, e.g., 'As a leader,

focusing on team synergy is critical,' aligns with the

tutor’s persona traits.

Equation 3: Tutor Selection and Response

Generation

The response from the AI tutor is generated using the

fine-tuned language model LM

as follows:



= LM



(Q) (3)

Where R

is the tutor's response and Q is the user's

question.

Each AI tutor's response R

depends on their training,

which captures both the knowledge domain and

personality traits.

3.4 Fine-Tuning Process

In the system, the external data feeds include

podcasts, speeches and social media posts that help to

infer the personality traits of the AI tutor (Balart, and,

Shryock, 2024). As shown in figure 3, the LLM is

fine-tuned using large amounts of data, including the

organized personality-specific data, to ensure it

behaves as designed. The fine-tuning process adjusts

LLMs parameters, so that it can mimic human like

interactions and respond to the personality traits

defined earlier in the system. This process ensures

that the interaction between the AI tutor and the user

continues to match the user expectations. The LLM is

fine tuned to match the exact traits necessary,

ensuring that the AI tutor will always behave in the

same way based on predefined persona. The fine-

tuning process is done by training each AI tutor to

their model over dataset that is specific to their

persona, adjusting its weight with the help of

supervised learning to anticipate the correct answer,

aligned with the tutor’s personality and domain

knowledge.

Equation 4: Fine-Tuning Loss Function

The loss function used during fine-tuning is

calculated as:

AI Driven Interactive Learning Platform: A Systematic Approach to Personalized Learning and Persona-Based Content Delivery

487

(

)

y



− f

(



,θ

)









(4)

Were,

• L(θ) is the loss,

• y

is the true label for response i,

• f(x

,θ) is the predicted response for input x

•

N is the number of training samples.

The process optimizes the model's parameters θ to

minimize this loss, improving the relevance and

accuracy of the AI tutor's responses.

Figure 3: Training and Fine-Tuning Pipeline

3.5 Content Generation and

Personalization

• Content Generation and Topic Search: There

is an initial “PROMPT” that creates content from

the user searches and course content accordingly.

This module employs topic search to get

pertinent contents for learning modules and

further, constructs learning content adaptively

based on the needs of the learner.

• Vectorization Using MiniLM Transformer:

The next step is vectorization of the generated

content, MiniLM transformer is used to obtain

the embedding (Virvou and Tsihrintzis, 2023).

These embeddings convert the textual data into

numerical signs, which make the knowledge

machine-readable. This is because it enables the

system to use the aspects of machine learning

models for personalization.

• Character-Based Content Personalization:

Another interesting feature within the system is

the option to adapt the contents depending on a

certain character of the person. In the “Course

Content + Character = Modified Content” block,

learning modules are transformed in terms of the

defined personas in order to enhance the

relevancy of the content for the learners.

•

Embedding Initialization and Training: The

generation of embeddings is based on the use of

the MiniLM transformer. These are the facts that

are learned from the course materials and from

other data sources through which the machine

learning models are trained. This phase helps to

personalized the content as well as to ensure that

the content presented is within the familiarity

level of the user and meets the intended learning

goals.

3.6 Learning Delivery

After selecting the tutor, the learning content is

delivered in modules as shown in figure 4. Each

module is presented as a conversation between the

user and the AI tutor(Nie and Nie, 2023), with the

tutor offering detailed explanations and clarifications.

Equation 5: Module Completion Check

At the end of each module, the system checks if the

user has understood the content using feedback-based

mechanisms. The completion score is calculated as:

C =

∑







(5)

Were,

• C is the completion score,

• u

is the user feedback for the i-th module,

• m is the total number of modules in the topic.

Figure 4: Learning Interface

INCOFT 2025 - International Conference on Futuristic Technology

488

3.7 Question-Answer Interaction

During learning session, users can follow up on

questions in real time. The system gives immediate

responses based on the fine-tuned model. The

interaction is dynamic, since the tutor is able to

change the way, they explain or answer questions in

order to capture the user’s attention (Gupta,

Dharamshi, et al. , 2024). The system is designed to

solve complex, context-dependent questions using

multi-turn dialogues. It also allows the conversation

to progress coherently across multiple user

interactions.

• Query Responses: The AI tutor powered by

LLM, accepts user query and processes it and

responds immediately having analyzed the user-

specific content made earlier. This interactive

module provides the ability to provide instant

feedback and answers in dual language to users

queries(Mohapatra, Shukla, et al. , 2018),

creating an engaging and real-time learning

experience.

• User Interaction Layer: The architecture

provides the necessary user interaction layer

where the user can search topic and ask queries.

The queries are then solved by the AI tutor,

which responds with personalized and

contextually relevant content (Kokku,

Sundararajan, et al. , 2018). It also allows for

immediate exchange which means that the

learning process and feedback is constant and

effectively makes the system very user-specific.

• Real-time Personalization: One of the standout

features of the system is its ability to perform

real-time personalization. As users interact with

the system, it dynamically adapts the course

content using the pre-defined characters. This

customization ensures that the learners engage

with the material in a more relatable and

immersive manner. The system modifies the

content’s tone, complexity and delivery style to

align with the chosen persona.

• Integration with External Learning

Materials: The system is designed to seamlessly

integrate external course materials in the

proposed framework. A number of materials

from different educational resources can be

input, analyze and transfer into vector form

based on the embedding model. The project

guarantees its compatibility with a wide range of

learning materials and thus remains quite flexible

and expandable where content type is concerned.

4 RESULTS AND DISCUSSION

In this section, we present the analysis of the

proposed AI-powered interactive learning platform’s

performance, in terms of accuracy, adaptability,

engagement and personalization across various AI

tutor personas. This evaluation metrics and outcomes

affirm the ability of the system to deliver an engaging

cognitive education environment.

Our system demonstrated a 30% improvement in

learner engagement compared to standard adaptive

systems, with an average session duration of 21

minutes (vs. 15 minutes in the baseline system) and a

response accuracy of 94%. These metrics highlight

the system’s potential for scalable deployment in

personalized learning environments.

4.1 Data Collection and Model Fine-

Tuning Results

Different dataset was used to train the AI tutor to

mimic particular tutor persona, such as Virat Kohli,

Doraemon, Steve Jobs and Shah Rukh Khan and so

on. Using data obtained from online resources,

conversations and user queries, it was possible to

perform further tuning of the LLM. After training, the

conformity of each AI persona to its respective

response was 94% and response generation time was

on average 1.2 seconds per query.

4.2 System Performance Analysis

The primary objective of this AI-powered system was

to deliver personalized, contextually accurate

responses that cater to user’s specific learning needs.

As shown in figure 5, the key metrics included

response accuracy, adaptability to user inputs, user

retention rates and average session duration. The

Table I, shows that Virat Kohli has the highest

Response Accuracy (96%) and Engagement Rate

(90%) with the shortest Average Response Time (1.1

seconds) and the longest Average Session Duration

(25 minutes). Doraemon follows closely in accuracy

(95%) but has a shorter session duration (18 minutes).

SRK and Steve Jobs perform moderately across these

metrics, with SRK showing a higher Engagement

Rate (88%) than Steve Jobs (83%). Overall, Virat

Kohli demonstrates the strongest performance in user

engagement and responsiveness.

AI Driven Interactive Learning Platform: A Systematic Approach to Personalized Learning and Persona-Based Content Delivery

489

Figure 5: Metrics Comparison Across Personas

Table 1: System Performance analysis Report

Metric Doraemon Virat

Kohli

SRK Steve

Jobs

Response

Accuracy

(

)

95 96 93 92

Avg.

Response

Time (s)

1.3 1.1 1.2 1.5

Avg.

Session

Duration

(

min

)

18 25 22 19

Engagement

Rate

(

)

85 90 88 83

4.3 Metrics for Accuracy Evaluation

• Response Accuracy: Manually review a random

sample of responses for relevance, accuracy and

alignment with each persona. Calculate the

proportion of correct or relevant responses for

each persona.

Response Acc.= 

Number of Relevant Responses

Total Responses



× 100 (6)

• Session Duration: Aggregate data on session

duration and frequency of interactions. Calculate

the average session duration for each persona.

Avg.Session Duration

=

Total Session Time for Persona

Total Number of Sessions

 (7)

• Engagement Rate: Calculate how many users

re-engage with the same persona. Engagement

rate can be represented by the percentage of

returning users.

Engagement Rate = 

Number of Returning Users

Total Unique Users



× 100 (8)

• Adaptability Accuracy: Track the system’s

ability to handle dynamically changing queries

by analyzing response relevance when user input

changes during a session.

Adaptability Acc.=

Relevant Dynamic Responses

Total Dynamic Queries



× 100 (9)

Figure 6: Performance Metrics – Avg. Across All Personas

The chart in figure 6, illustrates the performance

metrics of an AI-powered system, averaged across

different personas, highlighting its effectiveness and

user engagement. The system achieved a high

Engagement Rate of 86.50%, reflecting its ability to

capture and sustain user interest. An Average Session

Duration of 21 minutes demonstrates prolonged user

interaction, indicating the platform's relevance and

value. With an impressive Response Accuracy of

94%, the system consistently delivers reliable and

contextually appropriate answers. Additionally, a

swift Average Response Time of 1.28 seconds ensures

a seamless and efficient user experience. These

metrics collectively validate the system's potential for

providing an engaging and accurate platform for

users.

4.4 Error Rate and Resulting Accuracy

In order to examine the efficiency and the error rate

of the system for different personas, we performed

the evaluation for which we gained different error

INCOFT 2025 - International Conference on Futuristic Technology

490

rates because of the limitations in response generation

accuracy. As shown in the Table II, Doraemon had

8% error rate when answering in sports related tone

or responding with sports related answers thus

leading to a reduction in accuracy from 95% to 92%.

The responses given by Kohli was less erroneous

compared to other personas with an error percentage

of 5% with an accuracy reduction from 96% to 94%

was depicted. Steve Jobs and Shah Rukh Khan had a

slightly higher error rate of 9% and 7% bringing their

accuracy down to 88% and 90% from 92% & 93%

respectively. These findings suggest that refining

contextual filtering and enhancing persona-specific

knowledge could further improve response precision

across the system, ensuring relevant and accurate

replies aligned with each persona’s domain as

mentioned in the given figure 7 and figure 8.

Figure 7: Error Rate and Resulting Accuracy

Table 2: Evaluation Report

Persona

Erro

Rate

(%)

Reductio

n in

Accuracy

(%)

Actual

Respons

Accurac

y (%)

Resultin

Accurac

y (%)

Virat

Kohli

5 2 96 94

Doraemo

8 3 95 92

Steve

Jobs

9 4 92 88

Shah

Rukh

Khan

7 3 93 90

Avera

e 7.25 3 94 92

Figure 8: Average Error Rate and Resulting Accuracy

5 CONCLUSIONS

In Conclusion, this research demonstrates that a multi-

persona AI system suitable for delivering contextually

appropriated, nonrepetitive, accurate and engaging

responses based on the user’s personas is feasible. The

system has also demonstrated flexibility in tone and

mode of communication by adopting personalities like

Virat Kohli, Doraemon, Steve Jobs and Shah Rukh

Khan to increase the user satisfaction and engagement.

The performance data of the response accuracy,

session duration and engagement rates reveal that the

system performs well across the personas basically

proving that the system is a worthwhile investment

towards the improvement of other systems through the

utilization of user-centered AI. The outcomes

emphasize the benefits of persona approach to AI to

provide more natural and familiar communication

experience, which might be most effective in fields

such as customer support, education as well as

entertainment. This strategy’s future work could

involve work on extending persona possibilities, the

accuracy of contextual response and the question of

bias in order to build the strength of this scalable

model for a diverse range of AI-human interaction

scenarios. This study underscores the potential of

persona-driven AI in transforming education by

making learning more interactive and personalized.

Future research will focus on expanding cultural

diversity in personas and integrating multimodal

learning capabilities, such as video-based interaction.

AI Driven Interactive Learning Platform: A Systematic Approach to Personalized Learning and Persona-Based Content Delivery

491

REFERENCES

Rizvi, M. (2023). Investigating AI-Powered Tutoring

Systems that Adapt to Individual Student Needs,

Providing Personalized Guidance and Assessments.

The Eurasia Proceedings of Educational and Social

Sciences, 31, 67–73. doi.org/10.55549/epess.1381518.

Baillifard, A., Gabella, M., Lavenex, P. B., & Martarelli, C.

S. (2024). Effective learning with a personal AI tutor:

A case study. Education and Information Technologies.

doi.org/10.1007/s10639-024-12888-5.

G. Attigeri, A. Agrawal and S. V. Kolekar, "Advanced

NLP Models for Technical University Information

Chatbots: Development and Comparative Analysis," in

IEEE Access, vol. 12, pp. 29633-29647, 2024, doi:

10.1109/ACCESS.2024.3368382.

Arslan, B., Eyupoglu, G., Korkut, S., Turkdogan, K., &

Altinbilek, E. (2024). The accuracy of AI-assisted

chatbots on the annual assessment test for emergency

medicine residents. Journal of Medicine Surgery and

Public Health, 100070.

doi.org/10.1016/j.glmedi.2024.100070.

McGrath, C., Farazouli, A., & Cerratto-Pargman, T. (2024).

Generative AI chatbots in higher education: a review of

an emerging research area. Higher Education, doi:

10.1007/s10734-024-01288-w.

H. Shahri, M. Emad, N. Ibrahim, R. N. B. Rais and Y. Al-

Fayoumi, "Elevating Education through AI Tutor:

Utilizing GPT-4 for Personalized Learning," 2024 15th

Annual Undergraduate Research Conference on

Applied Computing (URC), Dubai, United Arab

Emirates, 2024, pp. 1-5, doi:

10.1109/URC62276.2024.10604578.

L. Frank, F. Herth, P. Stuwe, M. Klaiber, F. Gerschner and

A. Theissler, "Leveraging GenAI for an Intelligent

Tutoring System for R: A Quantitative Evaluation of

Large Language Models," 2024 IEEE Global

Engineering Education Conference (EDUCON), Kos

Island, Greece, 2024, pp. 1-9, doi:

10.1109/EDUCON60312.2024.10578933.

W. Gan, Y. Sun, S. Ye, Y. Fan and Y. Sun, "AI-Tutor:

Generating Tailored Remedial Questions and Answers

Based on Cognitive Diagnostic Assessment," 2019 6th

International Conference on Behavioral, Economic and

Socio-Cultural Computing (BESC), Beijing, China,

2019, pp. 1-6, doi:

10.1109/BESC48373.2019.8963236.

A. G, “RAG based Chatbot using LLMs,”

INTERANTIONAL JOURNAL OF SCIENTIFIC

RESEARCH IN ENGINEERING AND

MANAGEMENT, vol. 08, no. 06, pp. 1–5, Jun. 2024,

doi: 10.55041/ijsrem35600.

N. B. Trivedi, "AI in Education-A Transformative Force,"

2023 1st DMIHER International Conference on

Artificial Intelligence in Education and Industry 4.0

(IDICAIEI), Wardha, India, 2023, pp. 1-4, doi:

10.1109/IDICAIEI58380.2023.10406541.

T. Balart and K. J. Shryock, "Work in Progress:

Empowering Engineering Education With ChatGPT: A

Dive into the Potential and Challenges of Using AI for

Tutoring," 2024 IEEE Global Engineering Education

Conference (EDUCON), Kos Island, Greece, 2024, pp.

1-3, doi: 10.1109/EDUCON60312.2024.10578789.

S. Gupta, R. R. Dharamshi and V. Kakde, "An Impactful

and Revolutionized Educational Ecosystem using

Generative AI to Assist and Assess the Teaching and

Learning benefits, Fostering the Post-Pandemic

Requirements," 2024 Second International Conference

on Emerging Trends in Information Technology and

Engineering (ICETITE), Vellore, India, 2024, pp. 1-4,

doi: 10.1109/ic-ETITE58242.2024.10493370.

R. Makharia et al., "AI Tutor Enhanced with Prompt

Engineering and Deep Knowledge Tracing," 2024

IEEE International Conference on Interdisciplinary

Approaches in Technology and Management for Social

Innovation (IATMSI), Gwalior, India, 2024, pp. 1-6,

doi: 10.1109/IATMSI60426.2024.10503187.

F. AlShaikh and N. Hewahi, "AI and Machine Learning

Techniques in the Development of Intelligent Tutoring

System: A Review," 2021 International Conference on

Innovation and Intelligence for Informatics,

Computing, and Technologies (3ICT), Zallaq, Bahrain,

2021, pp. 403-410, doi:

10.1109/3ICT53449.2021.9582029.

R. Kokku, S. Sundararajan, P. Dey, R. Sindhgatta, S. Nitta

and B. Sengupta, "Augmenting Classrooms with AI for

Personalized Education," 2018 IEEE International

Conference on Acoustics, Speech and Signal

Processing (ICASSP), Calgary, AB, Canada, 2018, pp.

6976-6980, doi: 10.1109/ICASSP.2018.8461812.

M. Virvou and G. A. Tsihrintzis, "Is ChatGPT Beneficial to

Education? A Holistic Evaluation Framework Based on

Intelligent Tutoring Systems," 2023 14th International

Conference on Information, Intelligence, Systems &

Applications (IISA), Volos, Greece, 2023, pp. 1-8, doi:

10.1109/IISA59645.2023.10345949.

S. Mohapatra, N. Shukla, S. Jain and S. Chachra, "Nsmav-

Bot: Intelligent Dual Language Tutor System," 2018

Fourth International Conference on Computing

Communication Control and Automation (ICCUBEA),

Pune, India, 2018, pp. 1-5, doi:

10.1109/ICCUBEA.2018.8697582.

W. Shi, Z. Nie and Y. Shi, "Research on the Design and

Implementation of Intelligent Tutoring System Based

on AI Big Model," 2023 IEEE International Conference

on Unmanned Systems (ICUS), Hefei, China, 2023, pp.

1-6, doi: 10.1109/ICUS58632.2023.10318499.

I. Yesir and D. B. Rawat, "Recent Advances in Artificial

Intelligence Enabled Tutoring Systems: A Survey,"

2023 IEEE 13th Annual Computing and

Communication Workshop and Conference (CCWC),

Las Vegas, NV, USA, 2023, pp. 0375-0381, doi:

10.1109/CCWC57344.2023.10099098.

INCOFT 2025 - International Conference on Futuristic Technology

492