SMART CAREER ADVISOR : A MACHINE

LEARNING BASED RECOMMENDATION

SYSTEM

Asst. Prof. Rabecca Sandra Likitha G Suraj M

Dept of ISE Dept of ISE Dept of ISE
Bengaluru ,India Bengaluru , India Bengaluru , India
[email protected] [email protected] [email protected]

Jayanth R Monish V
Dept of ISE Dept of ISE
Bengaluru , India Bengaluru , India
[email protected] [email protected]

candidates to regularly update their resumes to meet current

Abstract—In the modern job market, candidates face standards.
significant challenges in securing roles that align with
their skills and interests due to the lack of personalized Recruiters increasingly rely on automated tools like
guidance and effective resume evaluation. This paper Applicant Tracking Systems (ATS) to efficiently filter
presents Career Craft AI, a personalized resume analysis resumes based on predefined criteria. To be shortlisted by
and job recommendation system integrated within an e- such systems, candidates must ensure their resumes align
learning platform. The system is designed to bridge the closely with job descriptions. However, manually analyzing
gap between candidates and suitable job opportunities resumes and matching them with relevant job roles is both
by analyzing user resumes using natural language time-consuming and prone to inaccuracies. To address these
processing techniques and identifying skill gaps through challenges, we propose a comprehensive solution: Career
machine learning. It further recommends relevant Craft AI. This platform not only helps users build ATS-
courses and certifications to improve user profiles. The friendly resumes but also analyzes resumes to recommend
job recommendation engine utilizes support vector skill enhancements, relevant courses, and personalized job
machines and cosine similarity to match users with opportunities.
suitable job roles based on their skills and experiences.
The system enhances resume quality to meet applicant Career Craft AI consists of two major components: a
tracking system standards and offers targeted job Resume Analyzer and a Job Recommendation System, both
suggestions, helping both candidates and recruiters. This integrated within an e-learning platform called Level Up.
approach aims to provide an intelligent, personalized, The system leverages natural language processing to extract
and effective solution for career development and job key resume features and assess their alignment with industry
placement. demands. Using support vector machine algorithms and
content-based filtering with cosine similarity, it clusters
Keywords—E-learning, recommendation system, content- users based on their skill profiles and recommends jobs
based filtering, natural language processing, support vector accordingly. The goal is to guide users in improving their
machine. resumes, developing in-demand skills, and increasing their
chances of securing roles that align with their professional
I. INTRODUCTION goals.

In today’s era of abundant career opportunities and evolving

job market dynamics, both fresh graduates and experienced
professionals often face challenges in identifying and II. LITERATURE REVIEW AND BACKGROUND
securing ideal job roles that match their unique skill sets and
aspirations. This difficulty is primarily due to an In recent years, the advancement of automated resume
overwhelming volume of information, a lack of personalized analysis and job recommendation systems has gained
job recommendations, and inadequate guidance on resume significant momentum due to the growing demand for
building. A well-structured and optimized resume plays a intelligent career guidance tools. Numerous platforms have
crucial role in the recruitment process, serving as the first emerged to assist candidates in enhancing their resumes by
impression for employers and significantly influencing offering features such as keyword optimization, formatting
shortlisting decisions. As job requirements and industry tips, and alignment with Applicant Tracking Systems (ATS).
expectations continue to change, it becomes essential for While these platforms improve resume visibility, many fall
short in addressing critical aspects such as skill gap analysis

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

and personalized career development. They rarely provide
targeted suggestions for certifications and training that align
with industry trends, thereby limiting their effectiveness in The diagram illustrates the complete workflow of the Career
preparing candidates for real-world job requirements. Craft AI system, depicting the interconnected stages
involved in providing personalized job recommendations
Several research studies have explored various techniques in and resume analysis. The process begins with data
the development of recommendation systems. Hyeyoung et collection, where relevant user information such as resumes,
al. conducted a survey that categorized recommendation skill profiles, and activity data from the e-learning platform
models based on their methodologies and fields of is gathered. This raw data is then passed through a data
application. Al-Otaibi et al. examined the structure of preprocessing phase, where inconsistencies, duplicates, and
existing online recruitment platforms, analyzing both the missing values are cleaned to ensure high-quality input for
strengths and limitations of different technical approaches. the model. Following this, feature engineering is conducted
Kethavarapu et al. introduced a job recommendation model to extract meaningful attributes from the processed data,
based on dynamic ontology created from job portal data, such as skill keywords, education details, and job
while Puspasari et al. implemented K-means clustering to preferences, which are then converted into structured
recommend jobs based on user preferences such as salary formats suitable for modeling.
and location. Appadoo et al. explored the integration of
machine learning and natural language processing for more The next step is model training, where machine learning
accurate job predictions. Other works, like those by Yu et al., algorithms, including Support Vector Machines (SVM), are
focused on resume parsing using cascaded hybrid models used to learn from the feature data and identify patterns for
for better information extraction. resume analysis and job matching. After training, the model
undergoes evaluation to assess its accuracy and
Most job recommendation systems adopt collaborative effectiveness using test data. Successful evaluation results
filtering, hybrid models, or content-based approaches. feed into the career database, which stores relevant job
Almalis et al. proposed a content-based model that enhances profiles and candidate insights. Based on the trained model,
traditional distance metrics to improve job-candidate recommendations are generated and delivered to the user
matching. Lu et al. combined content-based methods with a through an intuitive user interface, providing tailored job
PageRank-like algorithm to create a hybrid recommendation suggestions, resume improvement tips, and course
system. Popular platforms like LinkedIn and Glassdoor recommendations.
leverage these techniques to provide job suggestions based
on user profiles and interaction history. Despite their An essential component of the system is the feedback loop,
widespread use, these platforms often prioritize generalized where user interactions, preferences, and corrections are
suggestions over personalized ones based on detailed skill captured to continuously refine the recommendations. This
profiles and resume analysis. feedback is routed back to the data collection stage, forming
a continuous improvement cycle that enhances the
To address these gaps, Career Craft AI integrates both personalization and relevance of the system over time.
resume optimization and career development into a single
platform. By combining machine learning, natural language B. Modifications and Enhancements
processing, and personalized recommendation strategies, the
proposed system offers a more holistic approach to career
To improve the efficiency, accuracy, and user experience of
planning. It not only identifies and addresses skill gaps but
the Career Craft AI system, several modifications and
also suggests appropriate training and job opportunities enhancements have been integrated into the original model
tailored to each user, enhancing the overall effectiveness of architecture. First, the resume parsing module has been
job matching and candidate preparation.
enhanced with advanced Natural Language Processing (NLP)
techniques, enabling more accurate extraction of user
III METHODOLOGY information such as skills, experience, and education details.
This refinement ensures that the system captures nuanced
A. System Architecture and domain-specific terms, which improves the precision of
skill gap analysis and recommendations. Additionally, the
feature engineering process has been optimized to include
contextual embeddings using modern language models,
allowing the system to better understand the semantic
relevance between user skills and job descriptions.

To personalize recommendations further, the system now

incorporates user assessment performance data from the
integrated e-learning platform, adding an additional layer of
user profiling that enhances the accuracy of job and course
recommendations. The recommendation engine itself has
been upgraded with hybrid filtering techniques, combining
content-based filtering with behavioral insights gathered Vector Machine (SVM) and cosine similarity calculations.
from user interactions, leading to more relevant and Pandas and NumPy handle data manipulation and numerical
dynamic job suggestions. computations efficiently, while Matplotlib and Seaborn may
be used for data visualization during development and
The feedback loop mechanism has also been improved, evaluation phases. The system might also use Flask or
allowing real-time tracking of user engagement and Django as a web framework for integrating the modules into
capturing implicit feedback such as time spent on job the e-learning platform. Additionally, SQL or NoSQL
descriptions, course enrollments, and resume revision databases such as MySQL or MongoDB are used to store
patterns. This data is fed back into the system to retrain the user profiles, resume data, and job listings. These combined
model periodically, making it adaptive to user behavior and software tools and libraries form the backbone of the Career
evolving job market trends. Furthermore, the user interface Craft AI system, enabling seamless integration, high
has been enhanced with intuitive design elements, progress performance, and a personalized user experience.
tracking dashboards, and resume scoring indicators, making
the platform more interactive and user-friendly. C. Dataset

These strategic enhancements make the Career Craft AI The dataset used in the Career Craft AI system is a
system more robust, intelligent, and aligned with real-world combination of user-generated and system-collected data
recruitment needs, ultimately increasing its value to job tailored to support personalized resume analysis and job
seekers and recruiters alike. recommendations. It primarily consists of user profile data
gathered from the integrated e-learning platform, including
IV EXPERIMENTAL SETUP AND EVALUATION educational background, field of interest, existing skills,
career goals, and performance in quizzes and assessments.
A. Hardware Configuration In addition to this, the system utilizes resume data uploaded
by users, which is parsed using Natural Language
The hardware configuration required for implementing the Processing (NLP) techniques to extract structured
Career Craft AI system involves a robust setup capable of information such as work experience, skills, certifications,
efficiently handling tasks such as Natural Language and education. This data is then compared with industry-
Processing (NLP), machine learning model training, and standard skill requirements and job descriptions, likely
data processing. A system equipped with a multi-core sourced from publicly available job postings or curated
processor like Intel i7 or AMD Ryzen 7 (or server-grade company datasets. All collected data is stored in a
CPUs such as Intel Xeon) is recommended to manage centralized database and is used to train and fine-tune the
concurrent operations effectively. To support memory- recommendation model, particularly for identifying skill
intensive processes like resume parsing and model gaps and generating relevant job and course
execution, a minimum of 16 GB RAM is necessary, recommendations. The richness and relevance of this
although 32 GB or more is ideal. Solid State Drive (SSD) custom-built dataset ensure that the recommendations are
storage with at least 512 GB capacity ensures fast read/write accurate, personalized, and aligned with the dynamic needs
operations and sufficient space for storing resumes, user of the job market.
data, models, and related datasets. For systems employing
deep learning or complex NLP models, a dedicated NVIDIA
V RESULTS AND ANALYSIS
GPU with CUDA support, such as an RTX 3060 or higher,
is highly beneficial for accelerating computations. The
A Home Page
system can operate on platforms like Ubuntu 20.04 LTS or
Windows 10/11, depending on deployment needs. In
production environments, these specifications can be scaled
using cloud infrastructure services to enhance performance,
scalability, and reliability of the personalized resume
analysis and job recommendation functionalities.

B. Software Framework and Libraries

The software framework and libraries used in the Career

Craft AI system play a crucial role in enabling resume
analysis, skill gap identification, and personalized job
recommendations. The system is primarily built using
Python due to its versatility and strong ecosystem for data
science and machine learning. For Natural Language
Processing (NLP), libraries such as NLTK, spaCy, and
TextBlob are typically employed to parse and extract
relevant information from resumes. To implement the
recommendation engine and classification models, scikit-
learn is used for machine learning tasks including Support
B Login page resumes, typically within a few seconds. The use of Support
Vector Machine (SVM) for classification ensures efficient
computation, even with moderately large datasets, due to its
capability to handle high-dimensional data with minimal
overhead.

Feature extraction using techniques like TF-IDF and cosine

similarity is computationally efficient, enabling the system
to generate personalized job and skill recommendations
almost in real-time. Preprocessing and vectorization steps
are streamlined to minimize latency, and the system
architecture supports seamless integration of modules,
ensuring a smooth user experience.

Moreover, the recommendation engine is designed to be

scalable, allowing the system to maintain its performance
even as the number of users and resumes increases. This
ensures that the platform remains responsive and delivers
insights promptly, making it suitable for real-world
applications where time-sensitive career guidance and job
recommendations are critical.
C Quantitative Results
E Challenges and Limitation
The quantitative results of the Career Craft AI system
highlight its effectiveness in providing accurate and Despite the innovative approach and promising outcomes of
the Career Craft AI system, several challenges and
personalized resume analysis and job recommendations. In
limitations remain. One significant challenge lies in the
the evaluation phase, the system was tested using sample
accurate parsing and analysis of resumes, as the quality and
resumes sourced both from peers and publicly available
structure of uploaded documents may vary widely among
online templates. The results demonstrated that the system
could effectively identify relevant skills, detect skill gaps, users, potentially impacting the precision of skill extraction
and recommend appropriate courses and job roles. and analysis. Additionally, while the system uses machine
learning models like Support Vector Machines (SVM) and
content-based filtering for recommendations, the
As illustrated in the paper, a comparative analysis (shown in effectiveness of these models is highly dependent on the
Table I) between traditional systems and the proposed quality and completeness of the input data, which may not
model reveals that while traditional systems provide basic always be consistent or comprehensive. The
formatting guidance and ATS alignment, the Career Craft recommendation engine, although advanced, may still face
AI system excels by also offering skill gap analysis, course difficulties in identifying the nuanced preferences of users
and certification suggestions, and tailored job or adapting to rapidly evolving job market trends without
recommendations. The system’s ability to leverage Support frequent updates and retraining. Furthermore, the platform’s
Vector Machine (SVM) for classification and cosine reliance on cosine similarity and static skill-job mappings
similarity for profile matching enabled it to generate precise might limit its ability to handle complex job roles that
matches between user skills and job requirements. require a blend of interdisciplinary skills or soft skills that
are not easily quantifiable. Lastly, while integration with an
Additionally, the UI screenshots (Fig. 3–6) depict actionable e-learning platform provides a holistic experience, user
insights, including resume scoring, suggested improvements, engagement and motivation to follow the recommended
and a ranked list of jobs and courses. The resume scoring upskilling paths can vary, affecting the system's overall
mechanism effectively pinpointed weak areas and provided impact.
feedback to enhance user profiles. Based on the observed
testing outcomes, the system consistently delivered relevant VI CONCLUSION
and context-aware recommendations, indicating a high
degree of accuracy and practical applicability in real-world
job-matching scenarios. The Career Craft AI system offers a novel and robust
framework for addressing critical challenges in the job-
D Speed and Efficiency seeking process, particularly for individuals navigating an
increasingly competitive and skill-driven employment
The speed and efficiency of the Career Craft AI system are landscape. By combining advanced techniques from Natural
driven by its lightweight yet powerful machine learning and Language Processing (NLP), machine learning, and content-
Natural Language Processing (NLP) components. Resume based recommendation systems, the platform delivers a
parsing and skill extraction are performed using optimized unified solution for resume analysis, skill development, and
NLP libraries, which allow for quick processing of uploaded personalized job matching. The Resume Analyzer module
extracts and evaluates essential resume elements such as [5] K. Appadoo, M. B. Soonnoo, and Z. Mungloo-
education, experience, and skill sets, providing insightful Dilmohamud, "Job recommendation system, machine
feedback and identifying areas for improvement. learning, regression, classification, natural language
Simultaneously, the Job Recommendation System leverages processing," in Proc. 2020 IEEE Asia-Pacific Conf. Comput.
Support Vector Machine (SVM) classifiers and cosine Sci. Data Eng. (CSDE), 2020, pp. 1–6.
similarity algorithms to align user profiles with job
descriptions, ensuring recommendations are relevant and [6] K. Yu, G. Guan, and M. Zhou, "Resume information
tailored to individual competencies and aspirations. extraction with cascaded hybrid model," in Proc. 43rd Annu.
Meeting Assoc. Comput. Linguistics (ACL’05), 2005, pp.
What sets Career Craft AI apart from traditional systems is 499–506.
its integration with the e-learning platform, Level Up. This
allows the system to assess user performance through [7] N. D. Almalis, G. A. Tsihrintzis, N. Karagiannis, and A.
quizzes and coursework, thereby refining skill assessments D. Strati, "FoDRA—A new content-based job
and enhancing the accuracy of job recommendations. In recommendation algorithm for job seeking and recruiting,"
addition, the platform proactively identifies skill gaps and in Proc. 2015 6th Int. Conf. Inf., Intell., Syst. Appl. (IISA),
suggests targeted certifications and courses, thus supporting 2015, pp. 1–7.
continuous professional development and improving
candidates’ chances of being shortlisted by employers.
[8] Y. Lu, S. El Helou, and D. Gillet, "A recommender
system for job seeking and recruiting website," in Proc.
Through practical testing on real and sample resumes, the 22nd Int. Conf. World Wide Web, 2013, pp. 963–966.
system demonstrated effective performance in delivering
skill-based insights and relevant job suggestions, validating
[9] R. Aylward, "Glassdoor Expands Its Integrated Social
the underlying methodology. Moreover, the platform
Platform With Neo4j," Neo4j, 2014. [Online]. Available:
contributes to reducing the manual overhead for recruiters
http://info.neotechnology.com/rs/neotechnology/images/Gla
by pre-qualifying candidates with improved resumes aligned ssdoor.pdf
to industry standards.
[10] M. L. Tran, A. T. Nguyen, Q. D. Nguyen, and T.
Despite these achievements, the system also presents
Huynh, "A comparison study for job recommendation," in
opportunities for future enhancement, such as incorporating
Proc. 2017 Int. Conf. Inf. Commun. (ICIC), 2017, pp. 199–
more advanced AI models for deeper semantic 204.
understanding, extending support for multilingual resumes,
and integrating real-time labor market analytics to
dynamically update skill demands. Overall, Career Craft AI [11] M. Patel, "TinySearch—semantics based search engine
stands as a forward-thinking career guidance solution that using BERT embeddings," arXiv preprint arXiv:1908.02451,
not only empowers users to take control of their career paths 2019.
but also bridges the gap between talent and opportunity in a
data-driven, personalized manner. [12] C. Anastasiu et al., "DeepTitle—Leveraging BERT to
generate Search Engine Optimized Headlines," arXiv
preprint arXiv:2107.10935, 2021.

REFERENCES [13] I. S. Wahyudi, A. Affandi, and M. Hariadi,

"Recommender engine using cosine similarity based on
[1] H. Ko, S. Lee, Y. Park, and A. Choi, "A survey of alternating least square-weight regularization," in Proc.
recommendation systems: recommendation models, 2017 15th Int. Conf. Qual. Res. (QiR), 2017, pp. 256–261.
techniques, and application fields," Electronics, vol. 11, no.
1, p. 141, 2022. [14] S. Khapre, P. Singh, A. Shankar, S. R. Nayak, and M.
Diwakar, "Context-based intelligent recommendation by
[2] S. T. Al-Otaibi and M. Ykhlef, "A survey of job code reuse for smart decision support and cognitive adaptive
recommender systems," Int. J. Phys. Sci., vol. 7, no. 29, pp. systems," Int. J. Intell. Unmanned Syst., vol. 11, no. 1, pp.
5127–5142, 2012. 75–87, 2023.

[3] U. P. K. Kethavarapu and S. Saraswathi, "Concept based [15] S. Sharma et al., "Machine translation systems based on
dynamic ontology creation for job recommendation system," classical-statistical-deep-learning approaches," Electronics,
Procedia Comput. Sci., vol. 85, pp. 915–921, 2016. vol. 12, no. 7, p. 1716, 2023.

[4] B. D. Puspasari, L. L. Damayanti, A. Pramono, and A. K. [16] P. Singh et al., "Wireless sensor network-based
Darmawan, "Implementation K-means clustering method in monitoring system for health structure of rail-tracks: An
job recommendation system," in Proc. 2021 7th Int. Conf. efficient design for communication," 2023.
Electr., Electron. Inf. Eng. (ICEEIE), 2021, pp. 1–6.